1 libpng-manual.txt - A description on how to use and modify libpng
3 libpng version 1.6.13 - August 21, 2014
4 Updated and distributed by Glenn Randers-Pehrson
5 <glennrp at users.sourceforge.net>
6 Copyright (c) 1998-2014 Glenn Randers-Pehrson
8 This document is released under the libpng license.
9 For conditions of distribution and use, see the disclaimer
14 libpng versions 0.97, January 1998, through 1.6.13 - August 21, 2014
15 Updated and distributed by Glenn Randers-Pehrson
16 Copyright (c) 1998-2014 Glenn Randers-Pehrson
18 libpng 1.0 beta 6 - version 0.96 - May 28, 1997
19 Updated and distributed by Andreas Dilger
20 Copyright (c) 1996, 1997 Andreas Dilger
22 libpng 1.0 beta 2 - version 0.88 - January 26, 1996
23 For conditions of distribution and use, see copyright
24 notice in png.h. Copyright (c) 1995, 1996 Guy Eric
25 Schalnat, Group 42, Inc.
27 Updated/rewritten per request in the libpng FAQ
28 Copyright (c) 1995, 1996 Frank J. T. Wojcik
29 December 18, 1995 & January 20, 1996
38 VI. Modifying/Customizing libpng
40 VIII. Changes to Libpng from version 0.88
41 IX. Changes to Libpng from version 1.0.x to 1.2.x
42 X. Changes to Libpng from version 1.0.x/1.2.x to 1.4.x
43 XI. Changes to Libpng from version 1.4.x to 1.5.x
44 XII. Changes to Libpng from version 1.5.x to 1.6.x
45 XIII. Detecting libpng
46 XIV. Source code repository
48 XVI. Y2K Compliance in libpng
52 This file describes how to use and modify the PNG reference library
53 (known as libpng) for your own use. In addition to this
54 file, example.c is a good starting point for using the library, as
55 it is heavily commented and should include everything most people
56 will need. We assume that libpng is already installed; see the
57 INSTALL file for instructions on how to configure and install libpng.
59 For examples of libpng usage, see the files "example.c", "pngtest.c",
60 and the files in the "contrib" directory, all of which are included in
61 the libpng distribution.
63 Libpng was written as a companion to the PNG specification, as a way
64 of reducing the amount of time and effort it takes to support the PNG
65 file format in application programs.
67 The PNG specification (second edition), November 2003, is available as
68 a W3C Recommendation and as an ISO Standard (ISO/IEC 15948:2004 (E)) at
69 <http://www.w3.org/TR/2003/REC-PNG-20031110/
70 The W3C and ISO documents have identical technical content.
72 The PNG-1.2 specification is available at
73 <http://www.libpng.org/pub/png/documents/>. It is technically equivalent
74 to the PNG specification (second edition) but has some additional material.
76 The PNG-1.0 specification is available
77 as RFC 2083 <http://www.libpng.org/pub/png/documents/> and as a
78 W3C Recommendation <http://www.w3.org/TR/REC.png.html>.
80 Some additional chunks are described in the special-purpose public chunks
81 documents at <http://www.libpng.org/pub/png/documents/>.
84 about PNG, and the latest version of libpng, can be found at the PNG home
85 page, <http://www.libpng.org/pub/png/>.
87 Most users will not have to modify the library significantly; advanced
88 users may want to modify it more. All attempts were made to make it as
89 complete as possible, while keeping the code easy to understand.
90 Currently, this library only supports C. Support for other languages
93 Libpng has been designed to handle multiple sessions at one time,
94 to be easily modifiable, to be portable to the vast majority of
95 machines (ANSI, K&R, 16-, 32-, and 64-bit) available, and to be easy
96 to use. The ultimate goal of libpng is to promote the acceptance of
97 the PNG file format in whatever way possible. While there is still
98 work to be done (see the TODO file), libpng should cover the
99 majority of the needs of its users.
101 Libpng uses zlib for its compression and decompression of PNG files.
102 Further information about zlib, and the latest version of zlib, can
103 be found at the zlib home page, <http://www.info-zip.org/pub/infozip/zlib/>.
104 The zlib compression utility is a general purpose utility that is
105 useful for more than PNG files, and can be used without libpng.
106 See the documentation delivered with zlib for more details.
107 You can usually find the source files for the zlib utility wherever you
108 find the libpng source files.
110 Libpng is thread safe, provided the threads are using different
111 instances of the structures. Each thread should have its own
112 png_struct and png_info instances, and thus its own image.
113 Libpng does not protect itself against two threads using the
114 same instance of a structure.
118 There are two main structures that are important to libpng, png_struct
119 and png_info. Both are internal structures that are no longer exposed
120 in the libpng interface (as of libpng 1.5.0).
122 The png_info structure is designed to provide information about the
123 PNG file. At one time, the fields of png_info were intended to be
124 directly accessible to the user. However, this tended to cause problems
125 with applications using dynamically loaded libraries, and as a result
126 a set of interface functions for png_info (the png_get_*() and png_set_*()
127 functions) was developed, and direct access to the png_info fields was
130 The png_struct structure is the object used by the library to decode a
131 single image. As of 1.5.0 this structure is also not exposed.
133 Almost all libpng APIs require a pointer to a png_struct as the first argument.
134 Many (in particular the png_set and png_get APIs) also require a pointer
135 to png_info as the second argument. Some application visible macros
136 defined in png.h designed for basic data access (reading and writing
137 integers in the PNG format) don't take a png_info pointer, but it's almost
138 always safe to assume that a (png_struct*) has to be passed to call an API
141 You can have more than one png_info structure associated with an image,
142 as illustrated in pngtest.c, one for information valid prior to the
143 IDAT chunks and another (called "end_info" below) for things after them.
145 The png.h header file is an invaluable reference for programming with libpng.
146 And while I'm on the topic, make sure you include the libpng header file:
150 and also (as of libpng-1.5.0) the zlib header file, if you need it:
156 The png.h header file defines a number of integral types used by the
157 APIs. Most of these are fairly obvious; for example types corresponding
158 to integers of particular sizes and types for passing color values.
160 One exception is how non-integral numbers are handled. For application
161 convenience most APIs that take such numbers have C (double) arguments;
162 however, internally PNG, and libpng, use 32 bit signed integers and encode
163 the value by multiplying by 100,000. As of libpng 1.5.0 a convenience
164 macro PNG_FP_1 is defined in png.h along with a type (png_fixed_point)
165 which is simply (png_int_32).
167 All APIs that take (double) arguments also have a matching API that
168 takes the corresponding fixed point integer arguments. The fixed point
169 API has the same name as the floating point one with "_fixed" appended.
170 The actual range of values permitted in the APIs is frequently less than
171 the full range of (png_fixed_point) (-21474 to +21474). When APIs require
172 a non-negative argument the type is recorded as png_uint_32 above. Consult
173 the header file and the text below for more information.
175 Special care must be take with sCAL chunk handling because the chunk itself
176 uses non-integral values encoded as strings containing decimal floating point
177 numbers. See the comments in the header file.
181 The main header file function declarations are frequently protected by C
182 preprocessing directives of the form:
184 #ifdef PNG_feature_SUPPORTED
188 #ifdef PNG_feature_SUPPORTED
192 The library can be built without support for these APIs, although a
193 standard build will have all implemented APIs. Application programs
194 should check the feature macros before using an API for maximum
195 portability. From libpng 1.5.0 the feature macros set during the build
196 of libpng are recorded in the header file "pnglibconf.h" and this file
197 is always included by png.h.
199 If you don't need to change the library configuration from the default, skip to
200 the next section ("Reading").
202 Notice that some of the makefiles in the 'scripts' directory and (in 1.5.0) all
203 of the build project files in the 'projects' directory simply copy
204 scripts/pnglibconf.h.prebuilt to pnglibconf.h. This means that these build
205 systems do not permit easy auto-configuration of the library - they only
206 support the default configuration.
208 The easiest way to make minor changes to the libpng configuration when
209 auto-configuration is supported is to add definitions to the command line
210 using (typically) CPPFLAGS. For example:
212 CPPFLAGS=-DPNG_NO_FLOATING_ARITHMETIC
214 will change the internal libpng math implementation for gamma correction and
215 other arithmetic calculations to fixed point, avoiding the need for fast
216 floating point support. The result can be seen in the generated pnglibconf.h -
217 make sure it contains the changed feature macro setting.
219 If you need to make more extensive configuration changes - more than one or two
220 feature macro settings - you can either add -DPNG_USER_CONFIG to the build
221 command line and put a list of feature macro settings in pngusr.h or you can set
222 DFA_XTRA (a makefile variable) to a file containing the same information in the
223 form of 'option' settings.
225 A. Changing pnglibconf.h
227 A variety of methods exist to build libpng. Not all of these support
228 reconfiguration of pnglibconf.h. To reconfigure pnglibconf.h it must either be
229 rebuilt from scripts/pnglibconf.dfa using awk or it must be edited by hand.
231 Hand editing is achieved by copying scripts/pnglibconf.h.prebuilt to
232 pnglibconf.h and changing the lines defining the supported features, paying
233 very close attention to the 'option' information in scripts/pnglibconf.dfa
234 that describes those features and their requirements. This is easy to get
237 B. Configuration using DFA_XTRA
239 Rebuilding from pnglibconf.dfa is easy if a functioning 'awk', or a later
240 variant such as 'nawk' or 'gawk', is available. The configure build will
241 automatically find an appropriate awk and build pnglibconf.h.
242 The scripts/pnglibconf.mak file contains a set of make rules for doing the
243 same thing if configure is not used, and many of the makefiles in the scripts
244 directory use this approach.
246 When rebuilding simply write a new file containing changed options and set
247 DFA_XTRA to the name of this file. This causes the build to append the new file
248 to the end of scripts/pnglibconf.dfa. The pngusr.dfa file should contain lines
249 of the following forms:
253 This turns all optional features off. Include it at the start of pngusr.dfa to
254 make it easier to build a minimal configuration. You will need to turn at least
255 some features on afterward to enable either reading or writing code, or both.
260 Enable or disable a single feature. This will automatically enable other
261 features required by a feature that is turned on or disable other features that
262 require a feature which is turned off. Conflicting settings will cause an error
263 message to be emitted by awk.
265 setting feature default value
267 Changes the default value of setting 'feature' to 'value'. There are a small
268 number of settings listed at the top of pnglibconf.h, they are documented in the
269 source code. Most of these values have performance implications for the library
270 but most of them have no visible effect on the API. Some can also be overridden
273 This method of building a customized pnglibconf.h is illustrated in
274 contrib/pngminim/*. See the "$(PNGCONF):" target in the makefile and
275 pngusr.dfa in these directories.
277 C. Configuration using PNG_USER_CONFIG
279 If -DPNG_USER_CONFIG is added to the CPPFLAGS when pnglibconf.h is built,
280 the file pngusr.h will automatically be included before the options in
281 scripts/pnglibconf.dfa are processed. Your pngusr.h file should contain only
282 macro definitions turning features on or off or setting settings.
284 Apart from the global setting "everything = off" all the options listed above
285 can be set using macros in pngusr.h:
287 #define PNG_feature_SUPPORTED
293 #define PNG_NO_feature
299 #define PNG_feature value
303 setting feature default value
305 Notice that in both cases, pngusr.dfa and pngusr.h, the contents of the
306 pngusr file you supply override the contents of scripts/pnglibconf.dfa
308 If confusing or incomprehensible behavior results it is possible to
309 examine the intermediate file pnglibconf.dfn to find the full set of
310 dependency information for each setting and option. Simply locate the
311 feature in the file and read the C comments that precede it.
313 This method is also illustrated in the contrib/pngminim/* makefiles and
318 We'll now walk you through the possible functions to call when reading
319 in a PNG file sequentially, briefly explaining the syntax and purpose
320 of each one. See example.c and png.h for more detail. While
321 progressive reading is covered in the next section, you will still
322 need some of the functions discussed in this section to read a PNG
327 You will want to do the I/O initialization(*) before you get into libpng,
328 so if it doesn't work, you don't have much to undo. Of course, you
329 will also want to insure that you are, in fact, dealing with a PNG
330 file. Libpng provides a simple check to see if a file is a PNG file.
331 To use it, pass in the first 1 to 8 bytes of the file to the function
332 png_sig_cmp(), and it will return 0 (false) if the bytes match the
333 corresponding bytes of the PNG signature, or nonzero (true) otherwise.
334 Of course, the more bytes you pass in, the greater the accuracy of the
337 If you are intending to keep the file pointer open for use in libpng,
338 you must ensure you don't read more than 8 bytes from the beginning
339 of the file, and you also have to make a call to png_set_sig_bytes_read()
340 with the number of bytes you read from the beginning. Libpng will
341 then only check the bytes (if any) that your program didn't read.
343 (*): If you are not using the standard I/O functions, you will need
344 to replace them with custom functions. See the discussion under
348 FILE *fp = fopen(file_name, "rb");
354 fread(header, 1, number, fp);
355 is_png = !png_sig_cmp(header, 0, number);
363 Next, png_struct and png_info need to be allocated and initialized. In
364 order to ensure that the size of these structures is correct even with a
365 dynamically linked libpng, there are functions to initialize and
366 allocate the structures. We also pass the library version, optional
367 pointers to error handling functions, and a pointer to a data struct for
368 use by the error functions, if necessary (the pointer and functions can
369 be NULL if the default error handlers are to be used). See the section
370 on Changes to Libpng below regarding the old initialization functions.
371 The structure allocation functions quietly return NULL if they fail to
372 create the structure, so your application should check for that.
374 png_structp png_ptr = png_create_read_struct
375 (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
376 user_error_fn, user_warning_fn);
381 png_infop info_ptr = png_create_info_struct(png_ptr);
385 png_destroy_read_struct(&png_ptr,
386 (png_infopp)NULL, (png_infopp)NULL);
390 If you want to use your own memory allocation routines,
391 use a libpng that was built with PNG_USER_MEM_SUPPORTED defined, and use
392 png_create_read_struct_2() instead of png_create_read_struct():
394 png_structp png_ptr = png_create_read_struct_2
395 (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
396 user_error_fn, user_warning_fn, (png_voidp)
397 user_mem_ptr, user_malloc_fn, user_free_fn);
399 The error handling routines passed to png_create_read_struct()
400 and the memory alloc/free routines passed to png_create_struct_2()
401 are only necessary if you are not using the libpng supplied error
402 handling and memory alloc/free functions.
404 When libpng encounters an error, it expects to longjmp back
405 to your routine. Therefore, you will need to call setjmp and pass
406 your png_jmpbuf(png_ptr). If you read the file from different
407 routines, you will need to update the longjmp buffer every time you enter
408 a new routine that will call a png_*() function.
410 See your documentation of setjmp/longjmp for your compiler for more
411 information on setjmp/longjmp. See the discussion on libpng error
412 handling in the Customizing Libpng section below for more information
413 on the libpng error handling. If an error occurs, and libpng longjmp's
414 back to your setjmp, you will want to call png_destroy_read_struct() to
417 if (setjmp(png_jmpbuf(png_ptr)))
419 png_destroy_read_struct(&png_ptr, &info_ptr,
425 Pass (png_infopp)NULL instead of &end_info if you didn't create
426 an end_info structure.
428 If you would rather avoid the complexity of setjmp/longjmp issues,
429 you can compile libpng with PNG_NO_SETJMP, in which case
430 errors will result in a call to PNG_ABORT() which defaults to abort().
432 You can #define PNG_ABORT() to a function that does something
433 more useful than abort(), as long as your function does not
436 Now you need to set up the input code. The default for libpng is to
437 use the C function fread(). If you use this, you will need to pass a
438 valid FILE * in the function png_init_io(). Be sure that the file is
439 opened in binary mode. If you wish to handle reading data in another
440 way, you need not call the png_init_io() function, but you must then
441 implement the libpng I/O methods discussed in the Customizing Libpng
444 png_init_io(png_ptr, fp);
446 If you had previously opened the file and read any of the signature from
447 the beginning in order to see if this was a PNG file, you need to let
448 libpng know that there are some bytes missing from the start of the file.
450 png_set_sig_bytes(png_ptr, number);
452 You can change the zlib compression buffer size to be used while
453 reading compressed data with
455 png_set_compression_buffer_size(png_ptr, buffer_size);
457 where the default size is 8192 bytes. Note that the buffer size
458 is changed immediately and the buffer is reallocated immediately,
459 instead of setting a flag to be acted upon later.
461 If you want CRC errors to be handled in a different manner than
464 png_set_crc_action(png_ptr, crit_action, ancil_action);
466 The values for png_set_crc_action() say how libpng is to handle CRC errors in
467 ancillary and critical chunks, and whether to use the data contained
468 therein. Note that it is impossible to "discard" data in a critical
471 Choices for (int) crit_action are
472 PNG_CRC_DEFAULT 0 error/quit
473 PNG_CRC_ERROR_QUIT 1 error/quit
474 PNG_CRC_WARN_USE 3 warn/use data
475 PNG_CRC_QUIET_USE 4 quiet/use data
476 PNG_CRC_NO_CHANGE 5 use the current value
478 Choices for (int) ancil_action are
479 PNG_CRC_DEFAULT 0 error/quit
480 PNG_CRC_ERROR_QUIT 1 error/quit
481 PNG_CRC_WARN_DISCARD 2 warn/discard data
482 PNG_CRC_WARN_USE 3 warn/use data
483 PNG_CRC_QUIET_USE 4 quiet/use data
484 PNG_CRC_NO_CHANGE 5 use the current value
486 Setting up callback code
488 You can set up a callback function to handle any unknown chunks in the
489 input stream. You must supply the function
491 read_chunk_callback(png_structp png_ptr,
492 png_unknown_chunkp chunk);
494 /* The unknown chunk structure contains your
495 chunk data, along with similar data for any other
502 /* Note that libpng has already taken care of
505 /* put your code here. Search for your chunk in the
506 unknown chunk structure, process it, and return one
509 return (-n); /* chunk had an error */
510 return (0); /* did not recognize */
511 return (n); /* success */
514 (You can give your function another name that you like instead of
515 "read_chunk_callback")
517 To inform libpng about your function, use
519 png_set_read_user_chunk_fn(png_ptr, user_chunk_ptr,
520 read_chunk_callback);
522 This names not only the callback function, but also a user pointer that
523 you can retrieve with
525 png_get_user_chunk_ptr(png_ptr);
527 If you call the png_set_read_user_chunk_fn() function, then all unknown
528 chunks which the callback does not handle will be saved when read. You can
529 cause them to be discarded by returning '1' ("handled") instead of '0'. This
530 behavior will change in libpng 1.7 and the default handling set by the
531 png_set_keep_unknown_chunks() function, described below, will be used when the
532 callback returns 0. If you want the existing behavior you should set the global
533 default to PNG_HANDLE_CHUNK_IF_SAFE now; this is compatible with all current
534 versions of libpng and with 1.7. Libpng 1.6 issues a warning if you keep the
535 default, or PNG_HANDLE_CHUNK_NEVER, and the callback returns 0.
537 At this point, you can set up a callback function that will be
538 called after each row has been read, which you can use to control
539 a progress meter or the like. It's demonstrated in pngtest.c.
540 You must supply a function
542 void read_row_callback(png_structp png_ptr,
543 png_uint_32 row, int pass);
545 /* put your code here */
548 (You can give it another name that you like instead of "read_row_callback")
550 To inform libpng about your function, use
552 png_set_read_status_fn(png_ptr, read_row_callback);
554 When this function is called the row has already been completely processed and
555 the 'row' and 'pass' refer to the next row to be handled. For the
556 non-interlaced case the row that was just handled is simply one less than the
557 passed in row number, and pass will always be 0. For the interlaced case the
558 same applies unless the row value is 0, in which case the row just handled was
559 the last one from one of the preceding passes. Because interlacing may skip a
560 pass you cannot be sure that the preceding pass is just 'pass-1', if you really
561 need to know what the last pass is record (row,pass) from the callback and use
562 the last recorded value each time.
564 As with the user transform you can find the output row using the
565 PNG_ROW_FROM_PASS_ROW macro.
567 Unknown-chunk handling
569 Now you get to set the way the library processes unknown chunks in the
570 input PNG stream. Both known and unknown chunks will be read. Normal
571 behavior is that known chunks will be parsed into information in
572 various info_ptr members while unknown chunks will be discarded. This
573 behavior can be wasteful if your application will never use some known
574 chunk types. To change this, you can call:
576 png_set_keep_unknown_chunks(png_ptr, keep,
577 chunk_list, num_chunks);
579 keep - 0: default unknown chunk handling
580 1: ignore; do not keep
581 2: keep only if safe-to-copy
582 3: keep even if unsafe-to-copy
584 You can use these definitions:
585 PNG_HANDLE_CHUNK_AS_DEFAULT 0
586 PNG_HANDLE_CHUNK_NEVER 1
587 PNG_HANDLE_CHUNK_IF_SAFE 2
588 PNG_HANDLE_CHUNK_ALWAYS 3
590 chunk_list - list of chunks affected (a byte string,
591 five bytes per chunk, NULL or '\0' if
592 num_chunks is positive; ignored if
595 num_chunks - number of chunks affected; if 0, all
596 unknown chunks are affected. If positive,
597 only the chunks in the list are affected,
598 and if negative all unknown chunks and
599 all known chunks except for the IHDR,
600 PLTE, tRNS, IDAT, and IEND chunks are
603 Unknown chunks declared in this way will be saved as raw data onto a
604 list of png_unknown_chunk structures. If a chunk that is normally
605 known to libpng is named in the list, it will be handled as unknown,
606 according to the "keep" directive. If a chunk is named in successive
607 instances of png_set_keep_unknown_chunks(), the final instance will
608 take precedence. The IHDR and IEND chunks should not be named in
609 chunk_list; if they are, libpng will process them normally anyway.
610 If you know that your application will never make use of some particular
611 chunks, use PNG_HANDLE_CHUNK_NEVER (or 1) as demonstrated below.
613 Here is an example of the usage of png_set_keep_unknown_chunks(),
614 where the private "vpAg" chunk will later be processed by a user chunk
617 png_byte vpAg[5]={118, 112, 65, 103, (png_byte) '\0'};
619 #if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
620 png_byte unused_chunks[]=
622 104, 73, 83, 84, (png_byte) '\0', /* hIST */
623 105, 84, 88, 116, (png_byte) '\0', /* iTXt */
624 112, 67, 65, 76, (png_byte) '\0', /* pCAL */
625 115, 67, 65, 76, (png_byte) '\0', /* sCAL */
626 115, 80, 76, 84, (png_byte) '\0', /* sPLT */
627 116, 73, 77, 69, (png_byte) '\0', /* tIME */
633 #if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
634 /* ignore all unknown chunks
635 * (use global setting "2" for libpng16 and earlier):
637 png_set_keep_unknown_chunks(read_ptr, 2, NULL, 0);
639 /* except for vpAg: */
640 png_set_keep_unknown_chunks(read_ptr, 2, vpAg, 1);
642 /* also ignore unused known chunks: */
643 png_set_keep_unknown_chunks(read_ptr, 1, unused_chunks,
644 (int)(sizeof unused_chunks)/5);
649 The PNG specification allows the width and height of an image to be as
650 large as 2^31-1 (0x7fffffff), or about 2.147 billion rows and columns.
651 Since very few applications really need to process such large images,
652 we have imposed an arbitrary 1-million limit on rows and columns.
653 Larger images will be rejected immediately with a png_error() call. If
654 you wish to change this limit, you can use
656 png_set_user_limits(png_ptr, width_max, height_max);
658 to set your own limits, or use width_max = height_max = 0x7fffffffL
659 to allow all valid dimensions (libpng may reject some very large images
660 anyway because of potential buffer overflow conditions).
662 You should put this statement after you create the PNG structure and
663 before calling png_read_info(), png_read_png(), or png_process_data().
665 When writing a PNG datastream, put this statement before calling
666 png_write_info() or png_write_png().
668 If you need to retrieve the limits that are being applied, use
670 width_max = png_get_user_width_max(png_ptr);
671 height_max = png_get_user_height_max(png_ptr);
673 The PNG specification sets no limit on the number of ancillary chunks
674 allowed in a PNG datastream. You can impose a limit on the total number
675 of sPLT, tEXt, iTXt, zTXt, and unknown chunks that will be stored, with
677 png_set_chunk_cache_max(png_ptr, user_chunk_cache_max);
679 where 0x7fffffffL means unlimited. You can retrieve this limit with
681 chunk_cache_max = png_get_chunk_cache_max(png_ptr);
683 You can also set a limit on the amount of memory that a compressed chunk
684 other than IDAT can occupy, with
686 png_set_chunk_malloc_max(png_ptr, user_chunk_malloc_max);
688 and you can retrieve the limit with
690 chunk_malloc_max = png_get_chunk_malloc_max(png_ptr);
692 Any chunks that would cause either of these limits to be exceeded will
695 Information about your system
697 If you intend to display the PNG or to incorporate it in other image data you
698 need to tell libpng information about your display or drawing surface so that
699 libpng can convert the values in the image to match the display.
701 From libpng-1.5.4 this information can be set before reading the PNG file
702 header. In earlier versions png_set_gamma() existed but behaved incorrectly if
703 called before the PNG file header had been read and png_set_alpha_mode() did not
706 If you need to support versions prior to libpng-1.5.4 test the version number
707 as illustrated below using "PNG_LIBPNG_VER >= 10504" and follow the procedures
708 described in the appropriate manual page.
710 You give libpng the encoding expected by your system expressed as a 'gamma'
711 value. You can also specify a default encoding for the PNG file in
712 case the required information is missing from the file. By default libpng
713 assumes that the PNG data matches your system, to keep this default call:
715 png_set_gamma(png_ptr, screen_gamma, output_gamma);
717 or you can use the fixed point equivalent:
719 png_set_gamma_fixed(png_ptr, PNG_FP_1*screen_gamma,
720 PNG_FP_1*output_gamma);
722 If you don't know the gamma for your system it is probably 2.2 - a good
723 approximation to the IEC standard for display systems (sRGB). If images are
724 too contrasty or washed out you got the value wrong - check your system
727 Many systems permit the system gamma to be changed via a lookup table in the
728 display driver, a few systems, including older Macs, change the response by
729 default. As of 1.5.4 three special values are available to handle common
732 PNG_DEFAULT_sRGB: Indicates that the system conforms to the
733 IEC 61966-2-1 standard. This matches almost
735 PNG_GAMMA_MAC_18: Indicates that the system is an older
736 (pre Mac OS 10.6) Apple Macintosh system with
737 the default settings.
738 PNG_GAMMA_LINEAR: Just the fixed point value for 1.0 - indicates
739 that the system expects data with no gamma
742 You would use the linear (unencoded) value if you need to process the pixel
743 values further because this avoids the need to decode and re-encode each
744 component value whenever arithmetic is performed. A lot of graphics software
745 uses linear values for this reason, often with higher precision component values
746 to preserve overall accuracy.
749 The output_gamma value expresses how to decode the output values, not how
750 they are encoded. The values used correspond to the normal numbers used to
751 describe the overall gamma of a computer display system; for example 2.2 for
752 an sRGB conformant system. The values are scaled by 100000 in the _fixed
753 version of the API (so 220000 for sRGB.)
755 The inverse of the value is always used to provide a default for the PNG file
756 encoding if it has no gAMA chunk and if png_set_gamma() has not been called
757 to override the PNG gamma information.
759 When the ALPHA_OPTIMIZED mode is selected the output gamma is used to encode
760 opaque pixels however pixels with lower alpha values are not encoded,
761 regardless of the output gamma setting.
763 When the standard Porter Duff handling is requested with mode 1 the output
764 encoding is set to be linear and the output_gamma value is only relevant
765 as a default for input data that has no gamma information. The linear output
766 encoding will be overridden if png_set_gamma() is called - the results may be
769 The following numbers are derived from the sRGB standard and the research
770 behind it. sRGB is defined to be approximated by a PNG gAMA chunk value of
771 0.45455 (1/2.2) for PNG. The value implicitly includes any viewing
772 correction required to take account of any differences in the color
773 environment of the original scene and the intended display environment; the
774 value expresses how to *decode* the image for display, not how the original
777 sRGB provides a peg for the PNG standard by defining a viewing environment.
778 sRGB itself, and earlier TV standards, actually use a more complex transform
779 (a linear portion then a gamma 2.4 power law) than PNG can express. (PNG is
780 limited to simple power laws.) By saying that an image for direct display on
781 an sRGB conformant system should be stored with a gAMA chunk value of 45455
782 (11.3.3.2 and 11.3.3.5 of the ISO PNG specification) the PNG specification
783 makes it possible to derive values for other display systems and
786 The Mac value is deduced from the sRGB based on an assumption that the actual
787 extra viewing correction used in early Mac display systems was implemented as
788 a power 1.45 lookup table.
790 Any system where a programmable lookup table is used or where the behavior of
791 the final display device characteristics can be changed requires system
792 specific code to obtain the current characteristic. However this can be
793 difficult and most PNG gamma correction only requires an approximate value.
795 By default, if png_set_alpha_mode() is not called, libpng assumes that all
796 values are unencoded, linear, values and that the output device also has a
797 linear characteristic. This is only very rarely correct - it is invariably
798 better to call png_set_alpha_mode() with PNG_DEFAULT_sRGB than rely on the
799 default if you don't know what the right answer is!
801 The special value PNG_GAMMA_MAC_18 indicates an older Mac system (pre Mac OS
802 10.6) which used a correction table to implement a somewhat lower gamma on an
803 otherwise sRGB system.
805 Both these values are reserved (not simple gamma values) in order to allow
806 more precise correction internally in the future.
808 NOTE: the values can be passed to either the fixed or floating
809 point APIs, but the floating point API will also accept floating point
812 The second thing you may need to tell libpng about is how your system handles
813 alpha channel information. Some, but not all, PNG files contain an alpha
814 channel. To display these files correctly you need to compose the data onto a
815 suitable background, as described in the PNG specification.
817 Libpng only supports composing onto a single color (using png_set_background;
818 see below). Otherwise you must do the composition yourself and, in this case,
819 you may need to call png_set_alpha_mode:
821 #if PNG_LIBPNG_VER >= 10504
822 png_set_alpha_mode(png_ptr, mode, screen_gamma);
824 png_set_gamma(png_ptr, screen_gamma, 1.0/screen_gamma);
827 The screen_gamma value is the same as the argument to png_set_gamma; however,
828 how it affects the output depends on the mode. png_set_alpha_mode() sets the
829 file gamma default to 1/screen_gamma, so normally you don't need to call
830 png_set_gamma. If you need different defaults call png_set_gamma() before
831 png_set_alpha_mode() - if you call it after it will override the settings made
832 by png_set_alpha_mode().
834 The mode is as follows:
836 PNG_ALPHA_PNG: The data is encoded according to the PNG
837 specification. Red, green and blue, or gray, components are
838 gamma encoded color values and are not premultiplied by the
839 alpha value. The alpha value is a linear measure of the
840 contribution of the pixel to the corresponding final output pixel.
842 You should normally use this format if you intend to perform
843 color correction on the color values; most, maybe all, color
844 correction software has no handling for the alpha channel and,
845 anyway, the math to handle pre-multiplied component values is
846 unnecessarily complex.
848 Before you do any arithmetic on the component values you need
849 to remove the gamma encoding and multiply out the alpha
850 channel. See the PNG specification for more detail. It is
851 important to note that when an image with an alpha channel is
852 scaled, linear encoded, pre-multiplied component values must
855 The remaining modes assume you don't need to do any further color correction or
856 that if you do, your color correction software knows all about alpha (it
857 probably doesn't!). They 'associate' the alpha with the color information by
858 storing color channel values that have been scaled by the alpha. The
859 advantage is that the color channels can be resampled (the image can be
860 scaled) in this form. The disadvantage is that normal practice is to store
861 linear, not (gamma) encoded, values and this requires 16-bit channels for
862 still images rather than the 8-bit channels that are just about sufficient if
863 gamma encoding is used. In addition all non-transparent pixel values,
864 including completely opaque ones, must be gamma encoded to produce the final
865 image. These are the 'STANDARD', 'ASSOCIATED' or 'PREMULTIPLIED' modes
866 described below (the latter being the two common names for associated alpha
867 color channels). Note that PNG files always contain non-associated color
868 channels; png_set_alpha_mode() with one of the modes causes the decoder to
869 convert the pixels to an associated form before returning them to your
872 Since it is not necessary to perform arithmetic on opaque color values so
873 long as they are not to be resampled and are in the final color space it is
874 possible to optimize the handling of alpha by storing the opaque pixels in
875 the PNG format (adjusted for the output color space) while storing partially
876 opaque pixels in the standard, linear, format. The accuracy required for
877 standard alpha composition is relatively low, because the pixels are
878 isolated, therefore typically the accuracy loss in storing 8-bit linear
879 values is acceptable. (This is not true if the alpha channel is used to
880 simulate transparency over large areas - use 16 bits or the PNG mode in
881 this case!) This is the 'OPTIMIZED' mode. For this mode a pixel is
882 treated as opaque only if the alpha value is equal to the maximum value.
884 PNG_ALPHA_STANDARD: The data libpng produces is encoded in the
885 standard way assumed by most correctly written graphics software.
886 The gamma encoding will be removed by libpng and the
887 linear component values will be pre-multiplied by the
890 With this format the final image must be re-encoded to
891 match the display gamma before the image is displayed.
892 If your system doesn't do that, yet still seems to
893 perform arithmetic on the pixels without decoding them,
894 it is broken - check out the modes below.
896 With PNG_ALPHA_STANDARD libpng always produces linear
897 component values, whatever screen_gamma you supply. The
898 screen_gamma value is, however, used as a default for
899 the file gamma if the PNG file has no gamma information.
901 If you call png_set_gamma() after png_set_alpha_mode() you
902 will override the linear encoding. Instead the
903 pre-multiplied pixel values will be gamma encoded but
904 the alpha channel will still be linear. This may
905 actually match the requirements of some broken software,
908 While linear 8-bit data is often used it has
909 insufficient precision for any image with a reasonable
910 dynamic range. To avoid problems, and if your software
911 supports it, use png_set_expand_16() to force all
912 components to 16 bits.
914 PNG_ALPHA_OPTIMIZED: This mode is the same as PNG_ALPHA_STANDARD
915 except that completely opaque pixels are gamma encoded according to
916 the screen_gamma value. Pixels with alpha less than 1.0
917 will still have linear components.
919 Use this format if you have control over your
920 compositing software and so don't do other arithmetic
921 (such as scaling) on the data you get from libpng. Your
922 compositing software can simply copy opaque pixels to
923 the output but still has linear values for the
926 In normal compositing, where the alpha channel encodes
927 partial pixel coverage (as opposed to broad area
928 translucency), the inaccuracies of the 8-bit
929 representation of non-opaque pixels are irrelevant.
931 You can also try this format if your software is broken;
932 it might look better.
934 PNG_ALPHA_BROKEN: This is PNG_ALPHA_STANDARD; however, all component
935 values, including the alpha channel are gamma encoded. This is
936 broken because, in practice, no implementation that uses this choice
937 correctly undoes the encoding before handling alpha composition. Use this
938 choice only if other serious errors in the software or hardware you use
939 mandate it. In most cases of broken software or hardware the bug in the
940 final display manifests as a subtle halo around composited parts of the
941 image. You may not even perceive this as a halo; the composited part of
942 the image may simply appear separate from the background, as though it had
943 been cut out of paper and pasted on afterward.
945 If you don't have to deal with bugs in software or hardware, or if you can fix
946 them, there are three recommended ways of using png_set_alpha_mode():
948 png_set_alpha_mode(png_ptr, PNG_ALPHA_PNG,
951 You can do color correction on the result (libpng does not currently
952 support color correction internally). When you handle the alpha channel
953 you need to undo the gamma encoding and multiply out the alpha.
955 png_set_alpha_mode(png_ptr, PNG_ALPHA_STANDARD,
957 png_set_expand_16(png_ptr);
959 If you are using the high level interface, don't call png_set_expand_16();
960 instead pass PNG_TRANSFORM_EXPAND_16 to the interface.
962 With this mode you can't do color correction, but you can do arithmetic,
963 including composition and scaling, on the data without further processing.
965 png_set_alpha_mode(png_ptr, PNG_ALPHA_OPTIMIZED,
968 You can avoid the expansion to 16-bit components with this mode, but you
969 lose the ability to scale the image or perform other linear arithmetic.
970 All you can do is compose the result onto a matching output. Since this
971 mode is libpng-specific you also need to write your own composition
974 The following are examples of calls to png_set_alpha_mode to achieve the
975 required overall gamma correction and, where necessary, alpha
978 png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_DEFAULT_sRGB);
980 This is the default libpng handling of the alpha channel - it is not
981 pre-multiplied into the color components. In addition the call states
982 that the output is for a sRGB system and causes all PNG files without gAMA
983 chunks to be assumed to be encoded using sRGB.
985 png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_GAMMA_MAC);
987 In this case the output is assumed to be something like an sRGB conformant
988 display preceeded by a power-law lookup table of power 1.45. This is how
989 early Mac systems behaved.
991 png_set_alpha_mode(pp, PNG_ALPHA_STANDARD, PNG_GAMMA_LINEAR);
993 This is the classic Jim Blinn approach and will work in academic
994 environments where everything is done by the book. It has the shortcoming
995 of assuming that input PNG data with no gamma information is linear - this
996 is unlikely to be correct unless the PNG files where generated locally.
997 Most of the time the output precision will be so low as to show
998 significant banding in dark areas of the image.
1000 png_set_expand_16(pp);
1001 png_set_alpha_mode(pp, PNG_ALPHA_STANDARD, PNG_DEFAULT_sRGB);
1003 This is a somewhat more realistic Jim Blinn inspired approach. PNG files
1004 are assumed to have the sRGB encoding if not marked with a gamma value and
1005 the output is always 16 bits per component. This permits accurate scaling
1006 and processing of the data. If you know that your input PNG files were
1007 generated locally you might need to replace PNG_DEFAULT_sRGB with the
1008 correct value for your system.
1010 png_set_alpha_mode(pp, PNG_ALPHA_OPTIMIZED, PNG_DEFAULT_sRGB);
1012 If you just need to composite the PNG image onto an existing background
1013 and if you control the code that does this you can use the optimization
1014 setting. In this case you just copy completely opaque pixels to the
1015 output. For pixels that are not completely transparent (you just skip
1016 those) you do the composition math using png_composite or png_composite_16
1017 below then encode the resultant 8-bit or 16-bit values to match the output
1022 If neither the PNG nor the standard linear encoding work for you because
1023 of the software or hardware you use then you have a big problem. The PNG
1024 case will probably result in halos around the image. The linear encoding
1025 will probably result in a washed out, too bright, image (it's actually too
1026 contrasty.) Try the ALPHA_OPTIMIZED mode above - this will probably
1027 substantially reduce the halos. Alternatively try:
1029 png_set_alpha_mode(pp, PNG_ALPHA_BROKEN, PNG_DEFAULT_sRGB);
1031 This option will also reduce the halos, but there will be slight dark
1032 halos round the opaque parts of the image where the background is light.
1033 In the OPTIMIZED mode the halos will be light halos where the background
1034 is dark. Take your pick - the halos are unavoidable unless you can get
1035 your hardware/software fixed! (The OPTIMIZED approach is slightly
1038 When the default gamma of PNG files doesn't match the output gamma.
1039 If you have PNG files with no gamma information png_set_alpha_mode allows
1040 you to provide a default gamma, but it also sets the ouput gamma to the
1041 matching value. If you know your PNG files have a gamma that doesn't
1042 match the output you can take advantage of the fact that
1043 png_set_alpha_mode always sets the output gamma but only sets the PNG
1044 default if it is not already set:
1046 png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_DEFAULT_sRGB);
1047 png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_GAMMA_MAC);
1049 The first call sets both the default and the output gamma values, the
1050 second call overrides the output gamma without changing the default. This
1051 is easier than achieving the same effect with png_set_gamma. You must use
1052 PNG_ALPHA_PNG for the first call - internal checking in png_set_alpha will
1053 fire if more than one call to png_set_alpha_mode and png_set_background is
1054 made in the same read operation, however multiple calls with PNG_ALPHA_PNG
1057 If you don't need, or can't handle, the alpha channel you can call
1058 png_set_background() to remove it by compositing against a fixed color. Don't
1059 call png_set_strip_alpha() to do this - it will leave spurious pixel values in
1060 transparent parts of this image.
1062 png_set_background(png_ptr, &background_color,
1063 PNG_BACKGROUND_GAMMA_SCREEN, 0, 1);
1065 The background_color is an RGB or grayscale value according to the data format
1066 libpng will produce for you. Because you don't yet know the format of the PNG
1067 file, if you call png_set_background at this point you must arrange for the
1068 format produced by libpng to always have 8-bit or 16-bit components and then
1069 store the color as an 8-bit or 16-bit color as appropriate. The color contains
1070 separate gray and RGB component values, so you can let libpng produce gray or
1071 RGB output according to the input format, but low bit depth grayscale images
1072 must always be converted to at least 8-bit format. (Even though low bit depth
1073 grayscale images can't have an alpha channel they can have a transparent
1076 You set the transforms you need later, either as flags to the high level
1077 interface or libpng API calls for the low level interface. For reference the
1078 settings and API calls required are:
1081 PNG_TRANSFORM_SCALE_16 | PNG_EXPAND
1082 png_set_expand(png_ptr); png_set_scale_16(png_ptr);
1084 If you must get exactly the same inaccurate results
1085 produced by default in versions prior to libpng-1.5.4,
1086 use PNG_TRANSFORM_STRIP_16 and png_set_strip_16(png_ptr)
1090 PNG_TRANSFORM_EXPAND_16
1091 png_set_expand_16(png_ptr);
1093 In either case palette image data will be expanded to RGB. If you just want
1094 color data you can add PNG_TRANSFORM_GRAY_TO_RGB or png_set_gray_to_rgb(png_ptr)
1097 Calling png_set_background before the PNG file header is read will not work
1098 prior to libpng-1.5.4. Because the failure may result in unexpected warnings or
1099 errors it is therefore much safer to call png_set_background after the head has
1100 been read. Unfortunately this means that prior to libpng-1.5.4 it cannot be
1101 used with the high level interface.
1103 The high-level read interface
1105 At this point there are two ways to proceed; through the high-level
1106 read interface, or through a sequence of low-level read operations.
1107 You can use the high-level interface if (a) you are willing to read
1108 the entire image into memory, and (b) the input transformations
1109 you want to do are limited to the following set:
1111 PNG_TRANSFORM_IDENTITY No transformation
1112 PNG_TRANSFORM_SCALE_16 Strip 16-bit samples to
1114 PNG_TRANSFORM_STRIP_16 Chop 16-bit samples to
1115 8-bit less accurately
1116 PNG_TRANSFORM_STRIP_ALPHA Discard the alpha channel
1117 PNG_TRANSFORM_PACKING Expand 1, 2 and 4-bit
1119 PNG_TRANSFORM_PACKSWAP Change order of packed
1121 PNG_TRANSFORM_EXPAND Perform set_expand()
1122 PNG_TRANSFORM_INVERT_MONO Invert monochrome images
1123 PNG_TRANSFORM_SHIFT Normalize pixels to the
1125 PNG_TRANSFORM_BGR Flip RGB to BGR, RGBA
1127 PNG_TRANSFORM_SWAP_ALPHA Flip RGBA to ARGB or GA
1129 PNG_TRANSFORM_INVERT_ALPHA Change alpha from opacity
1131 PNG_TRANSFORM_SWAP_ENDIAN Byte-swap 16-bit samples
1132 PNG_TRANSFORM_GRAY_TO_RGB Expand grayscale samples
1133 to RGB (or GA to RGBA)
1134 PNG_TRANSFORM_EXPAND_16 Expand samples to 16 bits
1136 (This excludes setting a background color, doing gamma transformation,
1137 quantizing, and setting filler.) If this is the case, simply do this:
1139 png_read_png(png_ptr, info_ptr, png_transforms, NULL)
1141 where png_transforms is an integer containing the bitwise OR of some
1142 set of transformation flags. This call is equivalent to png_read_info(),
1143 followed the set of transformations indicated by the transform mask,
1144 then png_read_image(), and finally png_read_end().
1146 (The final parameter of this call is not yet used. Someday it might point
1147 to transformation parameters required by some future input transform.)
1149 You must use png_transforms and not call any png_set_transform() functions
1150 when you use png_read_png().
1152 After you have called png_read_png(), you can retrieve the image data
1155 row_pointers = png_get_rows(png_ptr, info_ptr);
1157 where row_pointers is an array of pointers to the pixel data for each row:
1159 png_bytep row_pointers[height];
1161 If you know your image size and pixel size ahead of time, you can allocate
1162 row_pointers prior to calling png_read_png() with
1164 if (height > PNG_UINT_32_MAX/(sizeof (png_byte)))
1166 "Image is too tall to process in memory");
1168 if (width > PNG_UINT_32_MAX/pixel_size)
1170 "Image is too wide to process in memory");
1172 row_pointers = png_malloc(png_ptr,
1173 height*(sizeof (png_bytep)));
1175 for (int i=0; i<height, i++)
1176 row_pointers[i]=NULL; /* security precaution */
1178 for (int i=0; i<height, i++)
1179 row_pointers[i]=png_malloc(png_ptr,
1182 png_set_rows(png_ptr, info_ptr, &row_pointers);
1184 Alternatively you could allocate your image in one big block and define
1185 row_pointers[i] to point into the proper places in your block.
1187 If you use png_set_rows(), the application is responsible for freeing
1188 row_pointers (and row_pointers[i], if they were separately allocated).
1190 If you don't allocate row_pointers ahead of time, png_read_png() will
1191 do it, and it'll be free'ed by libpng when you call png_destroy_*().
1193 The low-level read interface
1195 If you are going the low-level route, you are now ready to read all
1196 the file information up to the actual image data. You do this with a
1197 call to png_read_info().
1199 png_read_info(png_ptr, info_ptr);
1201 This will process all chunks up to but not including the image data.
1203 This also copies some of the data from the PNG file into the decode structure
1204 for use in later transformations. Important information copied in is:
1206 1) The PNG file gamma from the gAMA chunk. This overwrites the default value
1207 provided by an earlier call to png_set_gamma or png_set_alpha_mode.
1209 2) Prior to libpng-1.5.4 the background color from a bKGd chunk. This
1210 damages the information provided by an earlier call to png_set_background
1211 resulting in unexpected behavior. Libpng-1.5.4 no longer does this.
1213 3) The number of significant bits in each component value. Libpng uses this to
1214 optimize gamma handling by reducing the internal lookup table sizes.
1216 4) The transparent color information from a tRNS chunk. This can be modified by
1217 a later call to png_set_tRNS.
1219 Querying the info structure
1221 Functions are used to get the information from the info_ptr once it
1222 has been read. Note that these fields may not be completely filled
1223 in until png_read_end() has read the chunk data following the image.
1225 png_get_IHDR(png_ptr, info_ptr, &width, &height,
1226 &bit_depth, &color_type, &interlace_type,
1227 &compression_type, &filter_method);
1229 width - holds the width of the image
1230 in pixels (up to 2^31).
1232 height - holds the height of the image
1233 in pixels (up to 2^31).
1235 bit_depth - holds the bit depth of one of the
1236 image channels. (valid values are
1237 1, 2, 4, 8, 16 and depend also on
1238 the color_type. See also
1239 significant bits (sBIT) below).
1241 color_type - describes which color/alpha channels
1244 (bit depths 1, 2, 4, 8, 16)
1245 PNG_COLOR_TYPE_GRAY_ALPHA
1247 PNG_COLOR_TYPE_PALETTE
1248 (bit depths 1, 2, 4, 8)
1251 PNG_COLOR_TYPE_RGB_ALPHA
1254 PNG_COLOR_MASK_PALETTE
1255 PNG_COLOR_MASK_COLOR
1256 PNG_COLOR_MASK_ALPHA
1258 interlace_type - (PNG_INTERLACE_NONE or
1259 PNG_INTERLACE_ADAM7)
1261 compression_type - (must be PNG_COMPRESSION_TYPE_BASE
1264 filter_method - (must be PNG_FILTER_TYPE_BASE
1265 for PNG 1.0, and can also be
1266 PNG_INTRAPIXEL_DIFFERENCING if
1267 the PNG datastream is embedded in
1268 a MNG-1.0 datastream)
1270 Any or all of interlace_type, compression_type, or
1271 filter_method can be NULL if you are
1272 not interested in their values.
1274 Note that png_get_IHDR() returns 32-bit data into
1275 the application's width and height variables.
1276 This is an unsafe situation if these are 16-bit
1277 variables. In such situations, the
1278 png_get_image_width() and png_get_image_height()
1279 functions described below are safer.
1281 width = png_get_image_width(png_ptr,
1284 height = png_get_image_height(png_ptr,
1287 bit_depth = png_get_bit_depth(png_ptr,
1290 color_type = png_get_color_type(png_ptr,
1293 interlace_type = png_get_interlace_type(png_ptr,
1296 compression_type = png_get_compression_type(png_ptr,
1299 filter_method = png_get_filter_type(png_ptr,
1302 channels = png_get_channels(png_ptr, info_ptr);
1304 channels - number of channels of info for the
1305 color type (valid values are 1 (GRAY,
1306 PALETTE), 2 (GRAY_ALPHA), 3 (RGB),
1307 4 (RGB_ALPHA or RGB + filler byte))
1309 rowbytes = png_get_rowbytes(png_ptr, info_ptr);
1311 rowbytes - number of bytes needed to hold a row
1313 signature = png_get_signature(png_ptr, info_ptr);
1315 signature - holds the signature read from the
1316 file (if any). The data is kept in
1317 the same offset it would be if the
1318 whole signature were read (i.e. if an
1319 application had already read in 4
1320 bytes of signature before starting
1321 libpng, the remaining 4 bytes would
1322 be in signature[4] through signature[7]
1323 (see png_set_sig_bytes())).
1325 These are also important, but their validity depends on whether the chunk
1326 has been read. The png_get_valid(png_ptr, info_ptr, PNG_INFO_<chunk>) and
1327 png_get_<chunk>(png_ptr, info_ptr, ...) functions return non-zero if the
1328 data has been read, or zero if it is missing. The parameters to the
1329 png_get_<chunk> are set directly if they are simple data types, or a
1330 pointer into the info_ptr is returned for any complex types.
1332 The colorspace data from gAMA, cHRM, sRGB, iCCP, and sBIT chunks
1333 is simply returned to give the application information about how the
1334 image was encoded. Libpng itself only does transformations using the file
1335 gamma when combining semitransparent pixels with the background color, and,
1336 since libpng-1.6.0, when converting between 8-bit sRGB and 16-bit linear pixels
1337 within the simplified API. Libpng also uses the file gamma when converting
1338 RGB to gray, beginning with libpng-1.0.5, if the application calls
1339 png_set_rgb_to_gray()).
1341 png_get_PLTE(png_ptr, info_ptr, &palette,
1344 palette - the palette for the file
1345 (array of png_color)
1347 num_palette - number of entries in the palette
1349 png_get_gAMA(png_ptr, info_ptr, &file_gamma);
1350 png_get_gAMA_fixed(png_ptr, info_ptr, &int_file_gamma);
1352 file_gamma - the gamma at which the file is
1353 written (PNG_INFO_gAMA)
1355 int_file_gamma - 100,000 times the gamma at which the
1358 png_get_cHRM(png_ptr, info_ptr, &white_x, &white_y, &red_x,
1359 &red_y, &green_x, &green_y, &blue_x, &blue_y)
1360 png_get_cHRM_XYZ(png_ptr, info_ptr, &red_X, &red_Y, &red_Z,
1361 &green_X, &green_Y, &green_Z, &blue_X, &blue_Y,
1363 png_get_cHRM_fixed(png_ptr, info_ptr, &int_white_x,
1364 &int_white_y, &int_red_x, &int_red_y,
1365 &int_green_x, &int_green_y, &int_blue_x,
1367 png_get_cHRM_XYZ_fixed(png_ptr, info_ptr, &int_red_X, &int_red_Y,
1368 &int_red_Z, &int_green_X, &int_green_Y,
1369 &int_green_Z, &int_blue_X, &int_blue_Y,
1372 {white,red,green,blue}_{x,y}
1373 A color space encoding specified using the
1374 chromaticities of the end points and the
1375 white point. (PNG_INFO_cHRM)
1377 {red,green,blue}_{X,Y,Z}
1378 A color space encoding specified using the
1379 encoding end points - the CIE tristimulus
1380 specification of the intended color of the red,
1381 green and blue channels in the PNG RGB data.
1382 The white point is simply the sum of the three
1383 end points. (PNG_INFO_cHRM)
1385 png_get_sRGB(png_ptr, info_ptr, &srgb_intent);
1387 srgb_intent - the rendering intent (PNG_INFO_sRGB)
1388 The presence of the sRGB chunk
1389 means that the pixel data is in the
1390 sRGB color space. This chunk also
1391 implies specific values of gAMA and
1394 png_get_iCCP(png_ptr, info_ptr, &name,
1395 &compression_type, &profile, &proflen);
1397 name - The profile name.
1399 compression_type - The compression type; always
1400 PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
1401 You may give NULL to this argument to
1404 profile - International Color Consortium color
1405 profile data. May contain NULs.
1407 proflen - length of profile data in bytes.
1409 png_get_sBIT(png_ptr, info_ptr, &sig_bit);
1411 sig_bit - the number of significant bits for
1412 (PNG_INFO_sBIT) each of the gray,
1413 red, green, and blue channels,
1414 whichever are appropriate for the
1415 given color type (png_color_16)
1417 png_get_tRNS(png_ptr, info_ptr, &trans_alpha,
1418 &num_trans, &trans_color);
1420 trans_alpha - array of alpha (transparency)
1421 entries for palette (PNG_INFO_tRNS)
1423 num_trans - number of transparent entries
1426 trans_color - graylevel or color sample values of
1427 the single transparent color for
1428 non-paletted images (PNG_INFO_tRNS)
1430 png_get_hIST(png_ptr, info_ptr, &hist);
1433 hist - histogram of palette (array of
1436 png_get_tIME(png_ptr, info_ptr, &mod_time);
1438 mod_time - time image was last modified
1441 png_get_bKGD(png_ptr, info_ptr, &background);
1443 background - background color (of type
1444 png_color_16p) (PNG_VALID_bKGD)
1445 valid 16-bit red, green and blue
1446 values, regardless of color_type
1448 num_comments = png_get_text(png_ptr, info_ptr,
1449 &text_ptr, &num_text);
1451 num_comments - number of comments
1453 text_ptr - array of png_text holding image
1456 text_ptr[i].compression - type of compression used
1457 on "text" PNG_TEXT_COMPRESSION_NONE
1458 PNG_TEXT_COMPRESSION_zTXt
1459 PNG_ITXT_COMPRESSION_NONE
1460 PNG_ITXT_COMPRESSION_zTXt
1462 text_ptr[i].key - keyword for comment. Must contain
1465 text_ptr[i].text - text comments for current
1466 keyword. Can be empty.
1468 text_ptr[i].text_length - length of text string,
1469 after decompression, 0 for iTXt
1471 text_ptr[i].itxt_length - length of itxt string,
1472 after decompression, 0 for tEXt/zTXt
1474 text_ptr[i].lang - language of comment (empty
1475 string for unknown).
1477 text_ptr[i].lang_key - keyword in UTF-8
1478 (empty string for unknown).
1480 Note that the itxt_length, lang, and lang_key
1481 members of the text_ptr structure only exist when the
1482 library is built with iTXt chunk support. Prior to
1483 libpng-1.4.0 the library was built by default without
1484 iTXt support. Also note that when iTXt is supported,
1485 they contain NULL pointers when the "compression"
1486 field contains PNG_TEXT_COMPRESSION_NONE or
1487 PNG_TEXT_COMPRESSION_zTXt.
1489 num_text - number of comments (same as
1490 num_comments; you can put NULL here
1491 to avoid the duplication)
1493 Note while png_set_text() will accept text, language,
1494 and translated keywords that can be NULL pointers, the
1495 structure returned by png_get_text will always contain
1496 regular zero-terminated C strings. They might be
1497 empty strings but they will never be NULL pointers.
1499 num_spalettes = png_get_sPLT(png_ptr, info_ptr,
1502 num_spalettes - number of sPLT chunks read.
1504 palette_ptr - array of palette structures holding
1505 contents of one or more sPLT chunks
1508 png_get_oFFs(png_ptr, info_ptr, &offset_x, &offset_y,
1511 offset_x - positive offset from the left edge
1512 of the screen (can be negative)
1514 offset_y - positive offset from the top edge
1515 of the screen (can be negative)
1517 unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
1519 png_get_pHYs(png_ptr, info_ptr, &res_x, &res_y,
1522 res_x - pixels/unit physical resolution in
1525 res_y - pixels/unit physical resolution in
1528 unit_type - PNG_RESOLUTION_UNKNOWN,
1529 PNG_RESOLUTION_METER
1531 png_get_sCAL(png_ptr, info_ptr, &unit, &width,
1534 unit - physical scale units (an integer)
1536 width - width of a pixel in physical scale units
1538 height - height of a pixel in physical scale units
1539 (width and height are doubles)
1541 png_get_sCAL_s(png_ptr, info_ptr, &unit, &width,
1544 unit - physical scale units (an integer)
1546 width - width of a pixel in physical scale units
1547 (expressed as a string)
1549 height - height of a pixel in physical scale units
1550 (width and height are strings like "2.54")
1552 num_unknown_chunks = png_get_unknown_chunks(png_ptr,
1553 info_ptr, &unknowns)
1555 unknowns - array of png_unknown_chunk
1556 structures holding unknown chunks
1558 unknowns[i].name - name of unknown chunk
1560 unknowns[i].data - data of unknown chunk
1562 unknowns[i].size - size of unknown chunk's data
1564 unknowns[i].location - position of chunk in file
1566 The value of "i" corresponds to the order in which the
1567 chunks were read from the PNG file or inserted with the
1568 png_set_unknown_chunks() function.
1570 The value of "location" is a bitwise "or" of
1572 PNG_HAVE_IHDR (0x01)
1573 PNG_HAVE_PLTE (0x02)
1574 PNG_AFTER_IDAT (0x08)
1576 The data from the pHYs chunk can be retrieved in several convenient
1579 res_x = png_get_x_pixels_per_meter(png_ptr,
1582 res_y = png_get_y_pixels_per_meter(png_ptr,
1585 res_x_and_y = png_get_pixels_per_meter(png_ptr,
1588 res_x = png_get_x_pixels_per_inch(png_ptr,
1591 res_y = png_get_y_pixels_per_inch(png_ptr,
1594 res_x_and_y = png_get_pixels_per_inch(png_ptr,
1597 aspect_ratio = png_get_pixel_aspect_ratio(png_ptr,
1600 Each of these returns 0 [signifying "unknown"] if
1601 the data is not present or if res_x is 0;
1602 res_x_and_y is 0 if res_x != res_y
1604 Note that because of the way the resolutions are
1605 stored internally, the inch conversions won't
1606 come out to exactly even number. For example,
1607 72 dpi is stored as 0.28346 pixels/meter, and
1608 when this is retrieved it is 71.9988 dpi, so
1609 be sure to round the returned value appropriately
1610 if you want to display a reasonable-looking result.
1612 The data from the oFFs chunk can be retrieved in several convenient
1615 x_offset = png_get_x_offset_microns(png_ptr, info_ptr);
1617 y_offset = png_get_y_offset_microns(png_ptr, info_ptr);
1619 x_offset = png_get_x_offset_inches(png_ptr, info_ptr);
1621 y_offset = png_get_y_offset_inches(png_ptr, info_ptr);
1623 Each of these returns 0 [signifying "unknown" if both
1624 x and y are 0] if the data is not present or if the
1625 chunk is present but the unit is the pixel. The
1626 remark about inexact inch conversions applies here
1627 as well, because a value in inches can't always be
1628 converted to microns and back without some loss
1631 For more information, see the
1632 PNG specification for chunk contents. Be careful with trusting
1633 rowbytes, as some of the transformations could increase the space
1634 needed to hold a row (expand, filler, gray_to_rgb, etc.).
1635 See png_read_update_info(), below.
1637 A quick word about text_ptr and num_text. PNG stores comments in
1638 keyword/text pairs, one pair per chunk, with no limit on the number
1639 of text chunks, and a 2^31 byte limit on their size. While there are
1640 suggested keywords, there is no requirement to restrict the use to these
1641 strings. It is strongly suggested that keywords and text be sensible
1642 to humans (that's the point), so don't use abbreviations. Non-printing
1643 symbols are not allowed. See the PNG specification for more details.
1644 There is also no requirement to have text after the keyword.
1646 Keywords should be limited to 79 Latin-1 characters without leading or
1647 trailing spaces, but non-consecutive spaces are allowed within the
1648 keyword. It is possible to have the same keyword any number of times.
1649 The text_ptr is an array of png_text structures, each holding a
1650 pointer to a language string, a pointer to a keyword and a pointer to
1651 a text string. The text string, language code, and translated
1652 keyword may be empty or NULL pointers. The keyword/text
1653 pairs are put into the array in the order that they are received.
1654 However, some or all of the text chunks may be after the image, so, to
1655 make sure you have read all the text chunks, don't mess with these
1656 until after you read the stuff after the image. This will be
1657 mentioned again below in the discussion that goes with png_read_end().
1659 Input transformations
1661 After you've read the header information, you can set up the library
1662 to handle any special transformations of the image data. The various
1663 ways to transform the data will be described in the order that they
1664 should occur. This is important, as some of these change the color
1665 type and/or bit depth of the data, and some others only work on
1666 certain color types and bit depths.
1668 Transformations you request are ignored if they don't have any meaning for a
1669 particular input data format. However some transformations can have an effect
1670 as a result of a previous transformation. If you specify a contradictory set of
1671 transformations, for example both adding and removing the alpha channel, you
1672 cannot predict the final result.
1674 The color used for the transparency values should be supplied in the same
1675 format/depth as the current image data. It is stored in the same format/depth
1676 as the image data in a tRNS chunk, so this is what libpng expects for this data.
1678 The color used for the background value depends on the need_expand argument as
1681 Data will be decoded into the supplied row buffers packed into bytes
1682 unless the library has been told to transform it into another format.
1683 For example, 4 bit/pixel paletted or grayscale data will be returned
1684 2 pixels/byte with the leftmost pixel in the high-order bits of the
1685 byte, unless png_set_packing() is called. 8-bit RGB data will be stored
1686 in RGB RGB RGB format unless png_set_filler() or png_set_add_alpha()
1687 is called to insert filler bytes, either before or after each RGB triplet.
1688 16-bit RGB data will be returned RRGGBB RRGGBB, with the most significant
1689 byte of the color value first, unless png_set_scale_16() is called to
1690 transform it to regular RGB RGB triplets, or png_set_filler() or
1691 png_set_add alpha() is called to insert filler bytes, either before or
1692 after each RRGGBB triplet. Similarly, 8-bit or 16-bit grayscale data can
1693 be modified with png_set_filler(), png_set_add_alpha(), png_set_strip_16(),
1694 or png_set_scale_16().
1696 The following code transforms grayscale images of less than 8 to 8 bits,
1697 changes paletted images to RGB, and adds a full alpha channel if there is
1698 transparency information in a tRNS chunk. This is most useful on
1699 grayscale images with bit depths of 2 or 4 or if there is a multiple-image
1700 viewing application that wishes to treat all images in the same way.
1702 if (color_type == PNG_COLOR_TYPE_PALETTE)
1703 png_set_palette_to_rgb(png_ptr);
1705 if (png_get_valid(png_ptr, info_ptr,
1706 PNG_INFO_tRNS)) png_set_tRNS_to_alpha(png_ptr);
1708 if (color_type == PNG_COLOR_TYPE_GRAY &&
1709 bit_depth < 8) png_set_expand_gray_1_2_4_to_8(png_ptr);
1711 The first two functions are actually aliases for png_set_expand(), added
1712 in libpng version 1.0.4, with the function names expanded to improve code
1713 readability. In some future version they may actually do different
1716 As of libpng version 1.2.9, png_set_expand_gray_1_2_4_to_8() was
1717 added. It expands the sample depth without changing tRNS to alpha.
1719 As of libpng version 1.5.2, png_set_expand_16() was added. It behaves as
1720 png_set_expand(); however, the resultant channels have 16 bits rather than 8.
1721 Use this when the output color or gray channels are made linear to avoid fairly
1722 severe accuracy loss.
1725 png_set_expand_16(png_ptr);
1727 PNG can have files with 16 bits per channel. If you only can handle
1728 8 bits per channel, this will strip the pixels down to 8-bit.
1730 if (bit_depth == 16)
1731 #if PNG_LIBPNG_VER >= 10504
1732 png_set_scale_16(png_ptr);
1734 png_set_strip_16(png_ptr);
1737 (The more accurate "png_set_scale_16()" API became available in libpng version
1740 If you need to process the alpha channel on the image separately from the image
1741 data (for example if you convert it to a bitmap mask) it is possible to have
1742 libpng strip the channel leaving just RGB or gray data:
1744 if (color_type & PNG_COLOR_MASK_ALPHA)
1745 png_set_strip_alpha(png_ptr);
1747 If you strip the alpha channel you need to find some other way of dealing with
1748 the information. If, instead, you want to convert the image to an opaque
1749 version with no alpha channel use png_set_background; see below.
1751 As of libpng version 1.5.2, almost all useful expansions are supported, the
1752 major ommissions are conversion of grayscale to indexed images (which can be
1753 done trivially in the application) and conversion of indexed to grayscale (which
1754 can be done by a trivial manipulation of the palette.)
1756 In the following table, the 01 means grayscale with depth<8, 31 means
1757 indexed with depth<8, other numerals represent the color type, "T" means
1758 the tRNS chunk is present, A means an alpha channel is present, and O
1759 means tRNS or alpha is present but all pixels in the image are opaque.
1761 FROM 01 31 0 0T 0O 2 2T 2O 3 3T 3O 4A 4O 6A 6O
1763 01 - [G] - - - - - - - - - - - - -
1764 31 [Q] Q [Q] [Q] [Q] Q Q Q Q Q Q [Q] [Q] Q Q
1765 0 1 G + . . G G G G G G B B GB GB
1766 0T lt Gt t + . Gt G G Gt G G Bt Bt GBt GBt
1767 0O lt Gt t . + Gt Gt G Gt Gt G Bt Bt GBt GBt
1768 2 C P C C C + . . C - - CB CB B B
1769 2T Ct - Ct C C t + t - - - CBt CBt Bt Bt
1770 2O Ct - Ct C C t t + - - - CBt CBt Bt Bt
1771 3 [Q] p [Q] [Q] [Q] Q Q Q + . . [Q] [Q] Q Q
1772 3T [Qt] p [Qt][Q] [Q] Qt Qt Qt t + t [Qt][Qt] Qt Qt
1773 3O [Qt] p [Qt][Q] [Q] Qt Qt Qt t t + [Qt][Qt] Qt Qt
1774 4A lA G A T T GA GT GT GA GT GT + BA G GBA
1775 4O lA GBA A T T GA GT GT GA GT GT BA + GBA G
1776 6A CA PA CA C C A T tT PA P P C CBA + BA
1777 6O CA PBA CA C C A tT T PA P P CBA C BA +
1780 "+" identifies entries where 'from' and 'to' are the same.
1781 "-" means the transformation is not supported.
1782 "." means nothing is necessary (a tRNS chunk can just be ignored).
1783 "t" means the transformation is obtained by png_set_tRNS.
1784 "A" means the transformation is obtained by png_set_add_alpha().
1785 "X" means the transformation is obtained by png_set_expand().
1786 "1" means the transformation is obtained by
1787 png_set_expand_gray_1_2_4_to_8() (and by png_set_expand()
1788 if there is no transparency in the original or the final
1790 "C" means the transformation is obtained by png_set_gray_to_rgb().
1791 "G" means the transformation is obtained by png_set_rgb_to_gray().
1792 "P" means the transformation is obtained by
1793 png_set_expand_palette_to_rgb().
1794 "p" means the transformation is obtained by png_set_packing().
1795 "Q" means the transformation is obtained by png_set_quantize().
1796 "T" means the transformation is obtained by
1797 png_set_tRNS_to_alpha().
1798 "B" means the transformation is obtained by
1799 png_set_background(), or png_strip_alpha().
1801 When an entry has multiple transforms listed all are required to cause the
1802 right overall transformation. When two transforms are separated by a comma
1803 either will do the job. When transforms are enclosed in [] the transform should
1804 do the job but this is currently unimplemented - a different format will result
1805 if the suggested transformations are used.
1807 In PNG files, the alpha channel in an image
1808 is the level of opacity. If you need the alpha channel in an image to
1809 be the level of transparency instead of opacity, you can invert the
1810 alpha channel (or the tRNS chunk data) after it's read, so that 0 is
1811 fully opaque and 255 (in 8-bit or paletted images) or 65535 (in 16-bit
1812 images) is fully transparent, with
1814 png_set_invert_alpha(png_ptr);
1816 PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
1817 they can, resulting in, for example, 8 pixels per byte for 1 bit
1818 files. This code expands to 1 pixel per byte without changing the
1819 values of the pixels:
1822 png_set_packing(png_ptr);
1824 PNG files have possible bit depths of 1, 2, 4, 8, and 16. All pixels
1825 stored in a PNG image have been "scaled" or "shifted" up to the next
1826 higher possible bit depth (e.g. from 5 bits/sample in the range [0,31]
1827 to 8 bits/sample in the range [0, 255]). However, it is also possible
1828 to convert the PNG pixel data back to the original bit depth of the
1829 image. This call reduces the pixels back down to the original bit depth:
1831 png_color_8p sig_bit;
1833 if (png_get_sBIT(png_ptr, info_ptr, &sig_bit))
1834 png_set_shift(png_ptr, sig_bit);
1836 PNG files store 3-color pixels in red, green, blue order. This code
1837 changes the storage of the pixels to blue, green, red:
1839 if (color_type == PNG_COLOR_TYPE_RGB ||
1840 color_type == PNG_COLOR_TYPE_RGB_ALPHA)
1841 png_set_bgr(png_ptr);
1843 PNG files store RGB pixels packed into 3 or 6 bytes. This code expands them
1844 into 4 or 8 bytes for windowing systems that need them in this format:
1846 if (color_type == PNG_COLOR_TYPE_RGB)
1847 png_set_filler(png_ptr, filler, PNG_FILLER_BEFORE);
1849 where "filler" is the 8 or 16-bit number to fill with, and the location is
1850 either PNG_FILLER_BEFORE or PNG_FILLER_AFTER, depending upon whether
1851 you want the filler before the RGB or after. This transformation
1852 does not affect images that already have full alpha channels. To add an
1853 opaque alpha channel, use filler=0xff or 0xffff and PNG_FILLER_AFTER which
1854 will generate RGBA pixels.
1856 Note that png_set_filler() does not change the color type. If you want
1857 to do that, you can add a true alpha channel with
1859 if (color_type == PNG_COLOR_TYPE_RGB ||
1860 color_type == PNG_COLOR_TYPE_GRAY)
1861 png_set_add_alpha(png_ptr, filler, PNG_FILLER_AFTER);
1863 where "filler" contains the alpha value to assign to each pixel.
1864 This function was added in libpng-1.2.7.
1866 If you are reading an image with an alpha channel, and you need the
1867 data as ARGB instead of the normal PNG format RGBA:
1869 if (color_type == PNG_COLOR_TYPE_RGB_ALPHA)
1870 png_set_swap_alpha(png_ptr);
1872 For some uses, you may want a grayscale image to be represented as
1873 RGB. This code will do that conversion:
1875 if (color_type == PNG_COLOR_TYPE_GRAY ||
1876 color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
1877 png_set_gray_to_rgb(png_ptr);
1879 Conversely, you can convert an RGB or RGBA image to grayscale or grayscale
1882 if (color_type == PNG_COLOR_TYPE_RGB ||
1883 color_type == PNG_COLOR_TYPE_RGB_ALPHA)
1884 png_set_rgb_to_gray(png_ptr, error_action,
1885 double red_weight, double green_weight);
1887 error_action = 1: silently do the conversion
1889 error_action = 2: issue a warning if the original
1890 image has any pixel where
1891 red != green or red != blue
1893 error_action = 3: issue an error and abort the
1894 conversion if the original
1895 image has any pixel where
1896 red != green or red != blue
1898 red_weight: weight of red component
1900 green_weight: weight of green component
1901 If either weight is negative, default
1904 In the corresponding fixed point API the red_weight and green_weight values are
1905 simply scaled by 100,000:
1907 png_set_rgb_to_gray(png_ptr, error_action,
1908 png_fixed_point red_weight,
1909 png_fixed_point green_weight);
1911 If you have set error_action = 1 or 2, you can
1912 later check whether the image really was gray, after processing
1913 the image rows, with the png_get_rgb_to_gray_status(png_ptr) function.
1914 It will return a png_byte that is zero if the image was gray or
1915 1 if there were any non-gray pixels. Background and sBIT data
1916 will be silently converted to grayscale, using the green channel
1917 data for sBIT, regardless of the error_action setting.
1919 The default values come from the PNG file cHRM chunk if present; otherwise, the
1920 defaults correspond to the ITU-R recommendation 709, and also the sRGB color
1921 space, as recommended in the Charles Poynton's Colour FAQ,
1922 <http://www.poynton.com/>, in section 9:
1924 <http://www.poynton.com/notes/colour_and_gamma/ColorFAQ.html#RTFToC9>
1926 Y = 0.2126 * R + 0.7152 * G + 0.0722 * B
1928 Previous versions of this document, 1998 through 2002, recommended a slightly
1931 Y = 0.212671 * R + 0.715160 * G + 0.072169 * B
1933 Libpng uses an integer approximation:
1935 Y = (6968 * R + 23434 * G + 2366 * B)/32768
1937 The calculation is done in a linear colorspace, if the image gamma
1940 The png_set_background() function has been described already; it tells libpng to
1941 composite images with alpha or simple transparency against the supplied
1942 background color. For compatibility with versions of libpng earlier than
1943 libpng-1.5.4 it is recommended that you call the function after reading the file
1944 header, even if you don't want to use the color in a bKGD chunk, if one exists.
1946 If the PNG file contains a bKGD chunk (PNG_INFO_bKGD valid),
1947 you may use this color, or supply another color more suitable for
1948 the current display (e.g., the background color from a web page). You
1949 need to tell libpng how the color is represented, both the format of the
1950 component values in the color (the number of bits) and the gamma encoding of the
1951 color. The function takes two arguments, background_gamma_mode and need_expand
1952 to convey this information; however, only two combinations are likely to be
1955 png_color_16 my_background;
1956 png_color_16p image_background;
1958 if (png_get_bKGD(png_ptr, info_ptr, &image_background))
1959 png_set_background(png_ptr, image_background,
1960 PNG_BACKGROUND_GAMMA_FILE, 1/*needs to be expanded*/, 1);
1962 png_set_background(png_ptr, &my_background,
1963 PNG_BACKGROUND_GAMMA_SCREEN, 0/*do not expand*/, 1);
1965 The second call was described above - my_background is in the format of the
1966 final, display, output produced by libpng. Because you now know the format of
1967 the PNG it is possible to avoid the need to choose either 8-bit or 16-bit
1968 output and to retain palette images (the palette colors will be modified
1969 appropriately and the tRNS chunk removed.) However, if you are doing this,
1970 take great care not to ask for transformations without checking first that
1973 In the first call the background color has the original bit depth and color type
1974 of the PNG file. So, for palette images the color is supplied as a palette
1975 index and for low bit greyscale images the color is a reduced bit value in
1976 image_background->gray.
1978 If you didn't call png_set_gamma() before reading the file header, for example
1979 if you need your code to remain compatible with older versions of libpng prior
1980 to libpng-1.5.4, this is the place to call it.
1982 Do not call it if you called png_set_alpha_mode(); doing so will damage the
1983 settings put in place by png_set_alpha_mode(). (If png_set_alpha_mode() is
1984 supported then you can certainly do png_set_gamma() before reading the PNG
1987 This API unconditionally sets the screen and file gamma values, so it will
1988 override the value in the PNG file unless it is called before the PNG file
1989 reading starts. For this reason you must always call it with the PNG file
1990 value when you call it in this position:
1992 if (png_get_gAMA(png_ptr, info_ptr, &file_gamma))
1993 png_set_gamma(png_ptr, screen_gamma, file_gamma);
1996 png_set_gamma(png_ptr, screen_gamma, 0.45455);
1998 If you need to reduce an RGB file to a paletted file, or if a paletted
1999 file has more entries then will fit on your screen, png_set_quantize()
2000 will do that. Note that this is a simple match quantization that merely
2001 finds the closest color available. This should work fairly well with
2002 optimized palettes, but fairly badly with linear color cubes. If you
2003 pass a palette that is larger than maximum_colors, the file will
2004 reduce the number of colors in the palette so it will fit into
2005 maximum_colors. If there is a histogram, libpng will use it to make
2006 more intelligent choices when reducing the palette. If there is no
2007 histogram, it may not do as good a job.
2009 if (color_type & PNG_COLOR_MASK_COLOR)
2011 if (png_get_valid(png_ptr, info_ptr,
2014 png_uint_16p histogram = NULL;
2016 png_get_hIST(png_ptr, info_ptr,
2018 png_set_quantize(png_ptr, palette, num_palette,
2019 max_screen_colors, histogram, 1);
2024 png_color std_color_cube[MAX_SCREEN_COLORS] =
2027 png_set_quantize(png_ptr, std_color_cube,
2028 MAX_SCREEN_COLORS, MAX_SCREEN_COLORS,
2033 PNG files describe monochrome as black being zero and white being one.
2034 The following code will reverse this (make black be one and white be
2037 if (bit_depth == 1 && color_type == PNG_COLOR_TYPE_GRAY)
2038 png_set_invert_mono(png_ptr);
2040 This function can also be used to invert grayscale and gray-alpha images:
2042 if (color_type == PNG_COLOR_TYPE_GRAY ||
2043 color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
2044 png_set_invert_mono(png_ptr);
2046 PNG files store 16-bit pixels in network byte order (big-endian,
2047 ie. most significant bits first). This code changes the storage to the
2048 other way (little-endian, i.e. least significant bits first, the
2049 way PCs store them):
2051 if (bit_depth == 16)
2052 png_set_swap(png_ptr);
2054 If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
2055 need to change the order the pixels are packed into bytes, you can use:
2058 png_set_packswap(png_ptr);
2060 Finally, you can write your own transformation function if none of
2061 the existing ones meets your needs. This is done by setting a callback
2064 png_set_read_user_transform_fn(png_ptr,
2067 You must supply the function
2069 void read_transform_fn(png_structp png_ptr, png_row_infop
2070 row_info, png_bytep data)
2072 See pngtest.c for a working example. Your function will be called
2073 after all of the other transformations have been processed. Take care with
2074 interlaced images if you do the interlace yourself - the width of the row is the
2075 width in 'row_info', not the overall image width.
2077 If supported, libpng provides two information routines that you can use to find
2078 where you are in processing the image:
2080 png_get_current_pass_number(png_structp png_ptr);
2081 png_get_current_row_number(png_structp png_ptr);
2083 Don't try using these outside a transform callback - firstly they are only
2084 supported if user transforms are supported, secondly they may well return
2085 unexpected results unless the row is actually being processed at the moment they
2089 images the value returned is the row in the input sub-image image. Use
2090 PNG_ROW_FROM_PASS_ROW(row, pass) and PNG_COL_FROM_PASS_COL(col, pass) to
2091 find the output pixel (x,y) given an interlaced sub-image pixel (row,col,pass).
2093 The discussion of interlace handling above contains more information on how to
2096 You can also set up a pointer to a user structure for use by your
2097 callback function, and you can inform libpng that your transform
2098 function will change the number of channels or bit depth with the
2101 png_set_user_transform_info(png_ptr, user_ptr,
2102 user_depth, user_channels);
2104 The user's application, not libpng, is responsible for allocating and
2105 freeing any memory required for the user structure.
2107 You can retrieve the pointer via the function
2108 png_get_user_transform_ptr(). For example:
2110 voidp read_user_transform_ptr =
2111 png_get_user_transform_ptr(png_ptr);
2113 The last thing to handle is interlacing; this is covered in detail below,
2114 but you must call the function here if you want libpng to handle expansion
2115 of the interlaced image.
2117 number_of_passes = png_set_interlace_handling(png_ptr);
2119 After setting the transformations, libpng can update your png_info
2120 structure to reflect any transformations you've requested with this
2123 png_read_update_info(png_ptr, info_ptr);
2125 This is most useful to update the info structure's rowbytes
2126 field so you can use it to allocate your image memory. This function
2127 will also update your palette with the correct screen_gamma and
2128 background if these have been given with the calls above. You may
2129 only call png_read_update_info() once with a particular info_ptr.
2131 After you call png_read_update_info(), you can allocate any
2132 memory you need to hold the image. The row data is simply
2133 raw byte data for all forms of images. As the actual allocation
2134 varies among applications, no example will be given. If you
2135 are allocating one large chunk, you will need to build an
2136 array of pointers to each row, as it will be needed for some
2137 of the functions below.
2139 Remember: Before you call png_read_update_info(), the png_get_*()
2140 functions return the values corresponding to the original PNG image.
2141 After you call png_read_update_info the values refer to the image
2142 that libpng will output. Consequently you must call all the png_set_
2143 functions before you call png_read_update_info(). This is particularly
2144 important for png_set_interlace_handling() - if you are going to call
2145 png_read_update_info() you must call png_set_interlace_handling() before
2146 it unless you want to receive interlaced output.
2150 After you've allocated memory, you can read the image data.
2151 The simplest way to do this is in one function call. If you are
2152 allocating enough memory to hold the whole image, you can just
2153 call png_read_image() and libpng will read in all the image data
2154 and put it in the memory area supplied. You will need to pass in
2155 an array of pointers to each row.
2157 This function automatically handles interlacing, so you don't
2158 need to call png_set_interlace_handling() (unless you call
2159 png_read_update_info()) or call this function multiple times, or any
2160 of that other stuff necessary with png_read_rows().
2162 png_read_image(png_ptr, row_pointers);
2164 where row_pointers is:
2166 png_bytep row_pointers[height];
2168 You can point to void or char or whatever you use for pixels.
2170 If you don't want to read in the whole image at once, you can
2171 use png_read_rows() instead. If there is no interlacing (check
2172 interlace_type == PNG_INTERLACE_NONE), this is simple:
2174 png_read_rows(png_ptr, row_pointers, NULL,
2177 where row_pointers is the same as in the png_read_image() call.
2179 If you are doing this just one row at a time, you can do this with
2180 a single row_pointer instead of an array of row_pointers:
2182 png_bytep row_pointer = row;
2183 png_read_row(png_ptr, row_pointer, NULL);
2185 If the file is interlaced (interlace_type != 0 in the IHDR chunk), things
2186 get somewhat harder. The only current (PNG Specification version 1.2)
2187 interlacing type for PNG is (interlace_type == PNG_INTERLACE_ADAM7);
2188 a somewhat complicated 2D interlace scheme, known as Adam7, that
2189 breaks down an image into seven smaller images of varying size, based
2190 on an 8x8 grid. This number is defined (from libpng 1.5) as
2191 PNG_INTERLACE_ADAM7_PASSES in png.h
2193 libpng can fill out those images or it can give them to you "as is".
2194 It is almost always better to have libpng handle the interlacing for you.
2195 If you want the images filled out, there are two ways to do that. The one
2196 mentioned in the PNG specification is to expand each pixel to cover
2197 those pixels that have not been read yet (the "rectangle" method).
2198 This results in a blocky image for the first pass, which gradually
2199 smooths out as more pixels are read. The other method is the "sparkle"
2200 method, where pixels are drawn only in their final locations, with the
2201 rest of the image remaining whatever colors they were initialized to
2202 before the start of the read. The first method usually looks better,
2203 but tends to be slower, as there are more pixels to put in the rows.
2205 If, as is likely, you want libpng to expand the images, call this before
2206 calling png_start_read_image() or png_read_update_info():
2208 if (interlace_type == PNG_INTERLACE_ADAM7)
2210 = png_set_interlace_handling(png_ptr);
2212 This will return the number of passes needed. Currently, this is seven,
2213 but may change if another interlace type is added. This function can be
2214 called even if the file is not interlaced, where it will return one pass.
2215 You then need to read the whole image 'number_of_passes' times. Each time
2216 will distribute the pixels from the current pass to the correct place in
2217 the output image, so you need to supply the same rows to png_read_rows in
2220 If you are not going to display the image after each pass, but are
2221 going to wait until the entire image is read in, use the sparkle
2222 effect. This effect is faster and the end result of either method
2223 is exactly the same. If you are planning on displaying the image
2224 after each pass, the "rectangle" effect is generally considered the
2227 If you only want the "sparkle" effect, just call png_read_rows() as
2228 normal, with the third parameter NULL. Make sure you make pass over
2229 the image number_of_passes times, and you don't change the data in the
2230 rows between calls. You can change the locations of the data, just
2231 not the data. Each pass only writes the pixels appropriate for that
2232 pass, and assumes the data from previous passes is still valid.
2234 png_read_rows(png_ptr, row_pointers, NULL,
2237 If you only want the first effect (the rectangles), do the same as
2238 before except pass the row buffer in the third parameter, and leave
2239 the second parameter NULL.
2241 png_read_rows(png_ptr, NULL, row_pointers,
2244 If you don't want libpng to handle the interlacing details, just call
2245 png_read_rows() PNG_INTERLACE_ADAM7_PASSES times to read in all the images.
2246 Each of the images is a valid image by itself; however, you will almost
2247 certainly need to distribute the pixels from each sub-image to the
2248 correct place. This is where everything gets very tricky.
2250 If you want to retrieve the separate images you must pass the correct
2251 number of rows to each successive call of png_read_rows(). The calculation
2252 gets pretty complicated for small images, where some sub-images may
2253 not even exist because either their width or height ends up zero.
2254 libpng provides two macros to help you in 1.5 and later versions:
2256 png_uint_32 width = PNG_PASS_COLS(image_width, pass_number);
2257 png_uint_32 height = PNG_PASS_ROWS(image_height, pass_number);
2259 Respectively these tell you the width and height of the sub-image
2260 corresponding to the numbered pass. 'pass' is in in the range 0 to 6 -
2261 this can be confusing because the specification refers to the same passes
2262 as 1 to 7! Be careful, you must check both the width and height before
2263 calling png_read_rows() and not call it for that pass if either is zero.
2265 You can, of course, read each sub-image row by row. If you want to
2266 produce optimal code to make a pixel-by-pixel transformation of an
2267 interlaced image this is the best approach; read each row of each pass,
2268 transform it, and write it out to a new interlaced image.
2270 If you want to de-interlace the image yourself libpng provides further
2271 macros to help that tell you where to place the pixels in the output image.
2272 Because the interlacing scheme is rectangular - sub-image pixels are always
2273 arranged on a rectangular grid - all you need to know for each pass is the
2274 starting column and row in the output image of the first pixel plus the
2275 spacing between each pixel. As of libpng 1.5 there are four macros to
2276 retrieve this information:
2278 png_uint_32 x = PNG_PASS_START_COL(pass);
2279 png_uint_32 y = PNG_PASS_START_ROW(pass);
2280 png_uint_32 xStep = 1U << PNG_PASS_COL_SHIFT(pass);
2281 png_uint_32 yStep = 1U << PNG_PASS_ROW_SHIFT(pass);
2283 These allow you to write the obvious loop:
2285 png_uint_32 input_y = 0;
2286 png_uint_32 output_y = PNG_PASS_START_ROW(pass);
2288 while (output_y < output_image_height)
2290 png_uint_32 input_x = 0;
2291 png_uint_32 output_x = PNG_PASS_START_COL(pass);
2293 while (output_x < output_image_width)
2295 image[output_y][output_x] =
2296 subimage[pass][input_y][input_x++];
2305 Notice that the steps between successive output rows and columns are
2306 returned as shifts. This is possible because the pixels in the subimages
2307 are always a power of 2 apart - 1, 2, 4 or 8 pixels - in the original
2308 image. In practice you may need to directly calculate the output coordinate
2309 given an input coordinate. libpng provides two further macros for this
2312 png_uint_32 output_x = PNG_COL_FROM_PASS_COL(input_x, pass);
2313 png_uint_32 output_y = PNG_ROW_FROM_PASS_ROW(input_y, pass);
2315 Finally a pair of macros are provided to tell you if a particular image
2316 row or column appears in a given pass:
2318 int col_in_pass = PNG_COL_IN_INTERLACE_PASS(output_x, pass);
2319 int row_in_pass = PNG_ROW_IN_INTERLACE_PASS(output_y, pass);
2321 Bear in mind that you will probably also need to check the width and height
2322 of the pass in addition to the above to be sure the pass even exists!
2324 With any luck you are convinced by now that you don't want to do your own
2325 interlace handling. In reality normally the only good reason for doing this
2326 is if you are processing PNG files on a pixel-by-pixel basis and don't want
2327 to load the whole file into memory when it is interlaced.
2329 libpng includes a test program, pngvalid, that illustrates reading and
2330 writing of interlaced images. If you can't get interlacing to work in your
2331 code and don't want to leave it to libpng (the recommended approach), see
2332 how pngvalid.c does it.
2334 Finishing a sequential read
2336 After you are finished reading the image through the
2337 low-level interface, you can finish reading the file.
2339 If you want to use a different crc action for handling CRC errors in
2340 chunks after the image data, you can call png_set_crc_action()
2341 again at this point.
2343 If you are interested in comments or time, which may be stored either
2344 before or after the image data, you should pass the separate png_info
2345 struct if you want to keep the comments from before and after the image
2348 png_infop end_info = png_create_info_struct(png_ptr);
2352 png_destroy_read_struct(&png_ptr, &info_ptr,
2357 png_read_end(png_ptr, end_info);
2359 If you are not interested, you should still call png_read_end()
2360 but you can pass NULL, avoiding the need to create an end_info structure.
2361 If you do this, libpng will not process any chunks after IDAT other than
2362 skipping over them and perhaps (depending on whether you have called
2363 png_set_crc_action) checking their CRCs while looking for the IEND chunk.
2365 png_read_end(png_ptr, (png_infop)NULL);
2367 If you don't call png_read_end(), then your file pointer will be
2368 left pointing to the first chunk after the last IDAT, which is probably
2369 not what you want if you expect to read something beyond the end of
2372 When you are done, you can free all memory allocated by libpng like this:
2374 png_destroy_read_struct(&png_ptr, &info_ptr,
2377 or, if you didn't create an end_info structure,
2379 png_destroy_read_struct(&png_ptr, &info_ptr,
2382 It is also possible to individually free the info_ptr members that
2383 point to libpng-allocated storage with the following function:
2385 png_free_data(png_ptr, info_ptr, mask, seq)
2387 mask - identifies data to be freed, a mask
2388 containing the bitwise OR of one or
2390 PNG_FREE_PLTE, PNG_FREE_TRNS,
2391 PNG_FREE_HIST, PNG_FREE_ICCP,
2392 PNG_FREE_PCAL, PNG_FREE_ROWS,
2393 PNG_FREE_SCAL, PNG_FREE_SPLT,
2394 PNG_FREE_TEXT, PNG_FREE_UNKN,
2395 or simply PNG_FREE_ALL
2397 seq - sequence number of item to be freed
2400 This function may be safely called when the relevant storage has
2401 already been freed, or has not yet been allocated, or was allocated
2402 by the user and not by libpng, and will in those cases do nothing.
2403 The "seq" parameter is ignored if only one item of the selected data
2404 type, such as PLTE, is allowed. If "seq" is not -1, and multiple items
2405 are allowed for the data type identified in the mask, such as text or
2406 sPLT, only the n'th item in the structure is freed, where n is "seq".
2408 The default behavior is only to free data that was allocated internally
2409 by libpng. This can be changed, so that libpng will not free the data,
2410 or so that it will free data that was allocated by the user with png_malloc()
2411 or png_calloc() and passed in via a png_set_*() function, with
2413 png_data_freer(png_ptr, info_ptr, freer, mask)
2416 PNG_DESTROY_WILL_FREE_DATA
2417 PNG_SET_WILL_FREE_DATA
2418 PNG_USER_WILL_FREE_DATA
2420 mask - which data elements are affected
2421 same choices as in png_free_data()
2423 This function only affects data that has already been allocated.
2424 You can call this function after reading the PNG data but before calling
2425 any png_set_*() functions, to control whether the user or the png_set_*()
2426 function is responsible for freeing any existing data that might be present,
2427 and again after the png_set_*() functions to control whether the user
2428 or png_destroy_*() is supposed to free the data. When the user assumes
2429 responsibility for libpng-allocated data, the application must use
2430 png_free() to free it, and when the user transfers responsibility to libpng
2431 for data that the user has allocated, the user must have used png_malloc()
2432 or png_calloc() to allocate it.
2434 If you allocated your row_pointers in a single block, as suggested above in
2435 the description of the high level read interface, you must not transfer
2436 responsibility for freeing it to the png_set_rows or png_read_destroy function,
2437 because they would also try to free the individual row_pointers[i].
2439 If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
2440 separately, do not transfer responsibility for freeing text_ptr to libpng,
2441 because when libpng fills a png_text structure it combines these members with
2442 the key member, and png_free_data() will free only text_ptr.key. Similarly,
2443 if you transfer responsibility for free'ing text_ptr from libpng to your
2444 application, your application must not separately free those members.
2446 The png_free_data() function will turn off the "valid" flag for anything
2447 it frees. If you need to turn the flag off for a chunk that was freed by
2448 your application instead of by libpng, you can use
2450 png_set_invalid(png_ptr, info_ptr, mask);
2452 mask - identifies the chunks to be made invalid,
2453 containing the bitwise OR of one or
2455 PNG_INFO_gAMA, PNG_INFO_sBIT,
2456 PNG_INFO_cHRM, PNG_INFO_PLTE,
2457 PNG_INFO_tRNS, PNG_INFO_bKGD,
2458 PNG_INFO_hIST, PNG_INFO_pHYs,
2459 PNG_INFO_oFFs, PNG_INFO_tIME,
2460 PNG_INFO_pCAL, PNG_INFO_sRGB,
2461 PNG_INFO_iCCP, PNG_INFO_sPLT,
2462 PNG_INFO_sCAL, PNG_INFO_IDAT
2464 For a more compact example of reading a PNG image, see the file example.c.
2466 Reading PNG files progressively
2468 The progressive reader is slightly different from the non-progressive
2469 reader. Instead of calling png_read_info(), png_read_rows(), and
2470 png_read_end(), you make one call to png_process_data(), which calls
2471 callbacks when it has the info, a row, or the end of the image. You
2472 set up these callbacks with png_set_progressive_read_fn(). You don't
2473 have to worry about the input/output functions of libpng, as you are
2474 giving the library the data directly in png_process_data(). I will
2475 assume that you have read the section on reading PNG files above,
2476 so I will only highlight the differences (although I will show
2479 png_structp png_ptr;
2482 /* An example code fragment of how you would
2483 initialize the progressive reader in your
2486 initialize_png_reader()
2488 png_ptr = png_create_read_struct
2489 (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
2490 user_error_fn, user_warning_fn);
2495 info_ptr = png_create_info_struct(png_ptr);
2499 png_destroy_read_struct(&png_ptr,
2500 (png_infopp)NULL, (png_infopp)NULL);
2504 if (setjmp(png_jmpbuf(png_ptr)))
2506 png_destroy_read_struct(&png_ptr, &info_ptr,
2511 /* This one's new. You can provide functions
2512 to be called when the header info is valid,
2513 when each row is completed, and when the image
2514 is finished. If you aren't using all functions,
2515 you can specify NULL parameters. Even when all
2516 three functions are NULL, you need to call
2517 png_set_progressive_read_fn(). You can use
2518 any struct as the user_ptr (cast to a void pointer
2519 for the function call), and retrieve the pointer
2520 from inside the callbacks using the function
2522 png_get_progressive_ptr(png_ptr);
2524 which will return a void pointer, which you have
2525 to cast appropriately.
2527 png_set_progressive_read_fn(png_ptr, (void *)user_ptr,
2528 info_callback, row_callback, end_callback);
2533 /* A code fragment that you call as you receive blocks
2536 process_data(png_bytep buffer, png_uint_32 length)
2538 if (setjmp(png_jmpbuf(png_ptr)))
2540 png_destroy_read_struct(&png_ptr, &info_ptr,
2545 /* This one's new also. Simply give it a chunk
2546 of data from the file stream (in order, of
2547 course). On machines with segmented memory
2548 models machines, don't give it any more than
2549 64K. The library seems to run fine with sizes
2550 of 4K. Although you can give it much less if
2551 necessary (I assume you can give it chunks of
2552 1 byte, I haven't tried less then 256 bytes
2553 yet). When this function returns, you may
2554 want to display any rows that were generated
2555 in the row callback if you don't already do
2558 png_process_data(png_ptr, info_ptr, buffer, length);
2560 /* At this point you can call png_process_data_skip if
2561 you want to handle data the library will skip yourself;
2562 it simply returns the number of bytes to skip (and stops
2563 libpng skipping that number of bytes on the next
2564 png_process_data call).
2568 /* This function is called (as set by
2569 png_set_progressive_read_fn() above) when enough data
2570 has been supplied so all of the header has been
2574 info_callback(png_structp png_ptr, png_infop info)
2576 /* Do any setup here, including setting any of
2577 the transformations mentioned in the Reading
2578 PNG files section. For now, you _must_ call
2579 either png_start_read_image() or
2580 png_read_update_info() after all the
2581 transformations are set (even if you don't set
2582 any). You may start getting rows before
2583 png_process_data() returns, so this is your
2584 last chance to prepare for that.
2586 This is where you turn on interlace handling,
2587 assuming you don't want to do it yourself.
2589 If you need to you can stop the processing of
2590 your original input data at this point by calling
2591 png_process_data_pause. This returns the number
2592 of unprocessed bytes from the last png_process_data
2593 call - it is up to you to ensure that the next call
2594 sees these bytes again. If you don't want to bother
2595 with this you can get libpng to cache the unread
2596 bytes by setting the 'save' parameter (see png.h) but
2597 then libpng will have to copy the data internally.
2601 /* This function is called when each row of image
2604 row_callback(png_structp png_ptr, png_bytep new_row,
2605 png_uint_32 row_num, int pass)
2607 /* If the image is interlaced, and you turned
2608 on the interlace handler, this function will
2609 be called for every row in every pass. Some
2610 of these rows will not be changed from the
2611 previous pass. When the row is not changed,
2612 the new_row variable will be NULL. The rows
2613 and passes are called in order, so you don't
2614 really need the row_num and pass, but I'm
2615 supplying them because it may make your life
2618 If you did not turn on interlace handling then
2619 the callback is called for each row of each
2620 sub-image when the image is interlaced. In this
2621 case 'row_num' is the row in the sub-image, not
2622 the row in the output image as it is in all other
2625 For the non-NULL rows of interlaced images when
2626 you have switched on libpng interlace handling,
2627 you must call png_progressive_combine_row()
2628 passing in the row and the old row. You can
2629 call this function for NULL rows (it will just
2630 return) and for non-interlaced images (it just
2631 does the memcpy for you) if it will make the
2632 code easier. Thus, you can just do this for
2633 all cases if you switch on interlace handling;
2636 png_progressive_combine_row(png_ptr, old_row,
2639 /* where old_row is what was displayed
2640 previously for the row. Note that the first
2641 pass (pass == 0, really) will completely cover
2642 the old row, so the rows do not have to be
2643 initialized. After the first pass (and only
2644 for interlaced images), you will have to pass
2645 the current row, and the function will combine
2646 the old row and the new row.
2648 You can also call png_process_data_pause in this
2649 callback - see above.
2654 end_callback(png_structp png_ptr, png_infop info)
2656 /* This function is called after the whole image
2657 has been read, including any chunks after the
2658 image (up to and including the IEND). You
2659 will usually have the same info chunk as you
2660 had in the header, although some data may have
2661 been added to the comments and time fields.
2663 Most people won't do much here, perhaps setting
2664 a flag that marks the image as finished.
2672 Much of this is very similar to reading. However, everything of
2673 importance is repeated here, so you won't have to constantly look
2674 back up in the reading section to understand writing.
2678 You will want to do the I/O initialization before you get into libpng,
2679 so if it doesn't work, you don't have anything to undo. If you are not
2680 using the standard I/O functions, you will need to replace them with
2681 custom writing functions. See the discussion under Customizing libpng.
2683 FILE *fp = fopen(file_name, "wb");
2688 Next, png_struct and png_info need to be allocated and initialized.
2689 As these can be both relatively large, you may not want to store these
2690 on the stack, unless you have stack space to spare. Of course, you
2691 will want to check if they return NULL. If you are also reading,
2692 you won't want to name your read structure and your write structure
2693 both "png_ptr"; you can call them anything you like, such as
2694 "read_ptr" and "write_ptr". Look at pngtest.c, for example.
2696 png_structp png_ptr = png_create_write_struct
2697 (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
2698 user_error_fn, user_warning_fn);
2703 png_infop info_ptr = png_create_info_struct(png_ptr);
2706 png_destroy_write_struct(&png_ptr,
2711 If you want to use your own memory allocation routines,
2712 define PNG_USER_MEM_SUPPORTED and use
2713 png_create_write_struct_2() instead of png_create_write_struct():
2715 png_structp png_ptr = png_create_write_struct_2
2716 (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
2717 user_error_fn, user_warning_fn, (png_voidp)
2718 user_mem_ptr, user_malloc_fn, user_free_fn);
2720 After you have these structures, you will need to set up the
2721 error handling. When libpng encounters an error, it expects to
2722 longjmp() back to your routine. Therefore, you will need to call
2723 setjmp() and pass the png_jmpbuf(png_ptr). If you
2724 write the file from different routines, you will need to update
2725 the png_jmpbuf(png_ptr) every time you enter a new routine that will
2726 call a png_*() function. See your documentation of setjmp/longjmp
2727 for your compiler for more information on setjmp/longjmp. See
2728 the discussion on libpng error handling in the Customizing Libpng
2729 section below for more information on the libpng error handling.
2731 if (setjmp(png_jmpbuf(png_ptr)))
2733 png_destroy_write_struct(&png_ptr, &info_ptr);
2740 If you would rather avoid the complexity of setjmp/longjmp issues,
2741 you can compile libpng with PNG_NO_SETJMP, in which case
2742 errors will result in a call to PNG_ABORT() which defaults to abort().
2744 You can #define PNG_ABORT() to a function that does something
2745 more useful than abort(), as long as your function does not
2748 Checking for invalid palette index on write was added at libpng
2749 1.5.10. If a pixel contains an invalid (out-of-range) index libpng issues
2750 a benign error. This is enabled by default because this condition is an
2751 error according to the PNG specification, Clause 11.3.2, but the error can
2752 be ignored in each png_ptr with
2754 png_set_check_for_invalid_index(png_ptr, 0);
2756 If the error is ignored, or if png_benign_error() treats it as a warning,
2757 any invalid pixels are written as-is by the encoder, resulting in an
2758 invalid PNG datastream as output. In this case the application is
2759 responsible for ensuring that the pixel indexes are in range when it writes
2760 a PLTE chunk with fewer entries than the bit depth would allow.
2762 Now you need to set up the output code. The default for libpng is to
2763 use the C function fwrite(). If you use this, you will need to pass a
2764 valid FILE * in the function png_init_io(). Be sure that the file is
2765 opened in binary mode. Again, if you wish to handle writing data in
2766 another way, see the discussion on libpng I/O handling in the Customizing
2767 Libpng section below.
2769 png_init_io(png_ptr, fp);
2771 If you are embedding your PNG into a datastream such as MNG, and don't
2772 want libpng to write the 8-byte signature, or if you have already
2773 written the signature in your application, use
2775 png_set_sig_bytes(png_ptr, 8);
2777 to inform libpng that it should not write a signature.
2781 At this point, you can set up a callback function that will be
2782 called after each row has been written, which you can use to control
2783 a progress meter or the like. It's demonstrated in pngtest.c.
2784 You must supply a function
2786 void write_row_callback(png_structp png_ptr, png_uint_32 row,
2789 /* put your code here */
2792 (You can give it another name that you like instead of "write_row_callback")
2794 To inform libpng about your function, use
2796 png_set_write_status_fn(png_ptr, write_row_callback);
2798 When this function is called the row has already been completely processed and
2799 it has also been written out. The 'row' and 'pass' refer to the next row to be
2801 non-interlaced case the row that was just handled is simply one less than the
2802 passed in row number, and pass will always be 0. For the interlaced case the
2803 same applies unless the row value is 0, in which case the row just handled was
2804 the last one from one of the preceding passes. Because interlacing may skip a
2805 pass you cannot be sure that the preceding pass is just 'pass-1', if you really
2806 need to know what the last pass is record (row,pass) from the callback and use
2807 the last recorded value each time.
2809 As with the user transform you can find the output row using the
2810 PNG_ROW_FROM_PASS_ROW macro.
2812 You now have the option of modifying how the compression library will
2813 run. The following functions are mainly for testing, but may be useful
2814 in some cases, like if you need to write PNG files extremely fast and
2815 are willing to give up some compression, or if you want to get the
2816 maximum possible compression at the expense of slower writing. If you
2817 have no special needs in this area, let the library do what it wants by
2818 not calling this function at all, as it has been tuned to deliver a good
2819 speed/compression ratio. The second parameter to png_set_filter() is
2820 the filter method, for which the only valid values are 0 (as of the
2821 July 1999 PNG specification, version 1.2) or 64 (if you are writing
2822 a PNG datastream that is to be embedded in a MNG datastream). The third
2823 parameter is a flag that indicates which filter type(s) are to be tested
2824 for each scanline. See the PNG specification for details on the specific
2828 /* turn on or off filtering, and/or choose
2829 specific filters. You can use either a single
2830 PNG_FILTER_VALUE_NAME or the bitwise OR of one
2831 or more PNG_FILTER_NAME masks.
2833 png_set_filter(png_ptr, 0,
2834 PNG_FILTER_NONE | PNG_FILTER_VALUE_NONE |
2835 PNG_FILTER_SUB | PNG_FILTER_VALUE_SUB |
2836 PNG_FILTER_UP | PNG_FILTER_VALUE_UP |
2837 PNG_FILTER_AVG | PNG_FILTER_VALUE_AVG |
2838 PNG_FILTER_PAETH | PNG_FILTER_VALUE_PAETH|
2841 If an application wants to start and stop using particular filters during
2842 compression, it should start out with all of the filters (to ensure that
2843 the previous row of pixels will be stored in case it's needed later),
2844 and then add and remove them after the start of compression.
2846 If you are writing a PNG datastream that is to be embedded in a MNG
2847 datastream, the second parameter can be either 0 or 64.
2849 The png_set_compression_*() functions interface to the zlib compression
2850 library, and should mostly be ignored unless you really know what you are
2851 doing. The only generally useful call is png_set_compression_level()
2852 which changes how much time zlib spends on trying to compress the image
2853 data. See the Compression Library (zlib.h and algorithm.txt, distributed
2854 with zlib) for details on the compression levels.
2858 /* Set the zlib compression level */
2859 png_set_compression_level(png_ptr,
2860 Z_BEST_COMPRESSION);
2862 /* Set other zlib parameters for compressing IDAT */
2863 png_set_compression_mem_level(png_ptr, 8);
2864 png_set_compression_strategy(png_ptr,
2865 Z_DEFAULT_STRATEGY);
2866 png_set_compression_window_bits(png_ptr, 15);
2867 png_set_compression_method(png_ptr, 8);
2868 png_set_compression_buffer_size(png_ptr, 8192)
2870 /* Set zlib parameters for text compression
2871 * If you don't call these, the parameters
2872 * fall back on those defined for IDAT chunks
2874 png_set_text_compression_mem_level(png_ptr, 8);
2875 png_set_text_compression_strategy(png_ptr,
2876 Z_DEFAULT_STRATEGY);
2877 png_set_text_compression_window_bits(png_ptr, 15);
2878 png_set_text_compression_method(png_ptr, 8);
2880 Setting the contents of info for output
2882 You now need to fill in the png_info structure with all the data you
2883 wish to write before the actual image. Note that the only thing you
2884 are allowed to write after the image is the text chunks and the time
2885 chunk (as of PNG Specification 1.2, anyway). See png_write_end() and
2886 the latest PNG specification for more information on that. If you
2887 wish to write them before the image, fill them in now, and flag that
2888 data as being valid. If you want to wait until after the data, don't
2889 fill them until png_write_end(). For all the fields in png_info and
2890 their data types, see png.h. For explanations of what the fields
2891 contain, see the PNG specification.
2893 Some of the more important parts of the png_info are:
2895 png_set_IHDR(png_ptr, info_ptr, width, height,
2896 bit_depth, color_type, interlace_type,
2897 compression_type, filter_method)
2899 width - holds the width of the image
2900 in pixels (up to 2^31).
2902 height - holds the height of the image
2903 in pixels (up to 2^31).
2905 bit_depth - holds the bit depth of one of the
2907 (valid values are 1, 2, 4, 8, 16
2908 and depend also on the
2909 color_type. See also significant
2912 color_type - describes which color/alpha
2913 channels are present.
2915 (bit depths 1, 2, 4, 8, 16)
2916 PNG_COLOR_TYPE_GRAY_ALPHA
2918 PNG_COLOR_TYPE_PALETTE
2919 (bit depths 1, 2, 4, 8)
2922 PNG_COLOR_TYPE_RGB_ALPHA
2925 PNG_COLOR_MASK_PALETTE
2926 PNG_COLOR_MASK_COLOR
2927 PNG_COLOR_MASK_ALPHA
2929 interlace_type - PNG_INTERLACE_NONE or
2932 compression_type - (must be
2933 PNG_COMPRESSION_TYPE_DEFAULT)
2935 filter_method - (must be PNG_FILTER_TYPE_DEFAULT
2936 or, if you are writing a PNG to
2937 be embedded in a MNG datastream,
2939 PNG_INTRAPIXEL_DIFFERENCING)
2941 If you call png_set_IHDR(), the call must appear before any of the
2942 other png_set_*() functions, because they might require access to some of
2943 the IHDR settings. The remaining png_set_*() functions can be called
2946 If you wish, you can reset the compression_type, interlace_type, or
2947 filter_method later by calling png_set_IHDR() again; if you do this, the
2948 width, height, bit_depth, and color_type must be the same in each call.
2950 png_set_PLTE(png_ptr, info_ptr, palette,
2953 palette - the palette for the file
2954 (array of png_color)
2955 num_palette - number of entries in the palette
2957 png_set_gAMA(png_ptr, info_ptr, file_gamma);
2958 png_set_gAMA_fixed(png_ptr, info_ptr, int_file_gamma);
2960 file_gamma - the gamma at which the image was
2961 created (PNG_INFO_gAMA)
2963 int_file_gamma - 100,000 times the gamma at which
2964 the image was created
2966 png_set_cHRM(png_ptr, info_ptr, white_x, white_y, red_x, red_y,
2967 green_x, green_y, blue_x, blue_y)
2968 png_set_cHRM_XYZ(png_ptr, info_ptr, red_X, red_Y, red_Z, green_X,
2969 green_Y, green_Z, blue_X, blue_Y, blue_Z)
2970 png_set_cHRM_fixed(png_ptr, info_ptr, int_white_x, int_white_y,
2971 int_red_x, int_red_y, int_green_x, int_green_y,
2972 int_blue_x, int_blue_y)
2973 png_set_cHRM_XYZ_fixed(png_ptr, info_ptr, int_red_X, int_red_Y,
2974 int_red_Z, int_green_X, int_green_Y, int_green_Z,
2975 int_blue_X, int_blue_Y, int_blue_Z)
2977 {white,red,green,blue}_{x,y}
2978 A color space encoding specified using the chromaticities
2979 of the end points and the white point.
2981 {red,green,blue}_{X,Y,Z}
2982 A color space encoding specified using the encoding end
2983 points - the CIE tristimulus specification of the intended
2984 color of the red, green and blue channels in the PNG RGB
2985 data. The white point is simply the sum of the three end
2988 png_set_sRGB(png_ptr, info_ptr, srgb_intent);
2990 srgb_intent - the rendering intent
2991 (PNG_INFO_sRGB) The presence of
2992 the sRGB chunk means that the pixel
2993 data is in the sRGB color space.
2994 This chunk also implies specific
2995 values of gAMA and cHRM. Rendering
2996 intent is the CSS-1 property that
2997 has been defined by the International
2999 (http://www.color.org).
3001 PNG_sRGB_INTENT_SATURATION,
3002 PNG_sRGB_INTENT_PERCEPTUAL,
3003 PNG_sRGB_INTENT_ABSOLUTE, or
3004 PNG_sRGB_INTENT_RELATIVE.
3007 png_set_sRGB_gAMA_and_cHRM(png_ptr, info_ptr,
3010 srgb_intent - the rendering intent
3011 (PNG_INFO_sRGB) The presence of the
3012 sRGB chunk means that the pixel
3013 data is in the sRGB color space.
3014 This function also causes gAMA and
3015 cHRM chunks with the specific values
3016 that are consistent with sRGB to be
3019 png_set_iCCP(png_ptr, info_ptr, name, compression_type,
3022 name - The profile name.
3024 compression_type - The compression type; always
3025 PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
3026 You may give NULL to this argument to
3029 profile - International Color Consortium color
3030 profile data. May contain NULs.
3032 proflen - length of profile data in bytes.
3034 png_set_sBIT(png_ptr, info_ptr, sig_bit);
3036 sig_bit - the number of significant bits for
3037 (PNG_INFO_sBIT) each of the gray, red,
3038 green, and blue channels, whichever are
3039 appropriate for the given color type
3042 png_set_tRNS(png_ptr, info_ptr, trans_alpha,
3043 num_trans, trans_color);
3045 trans_alpha - array of alpha (transparency)
3046 entries for palette (PNG_INFO_tRNS)
3048 num_trans - number of transparent entries
3051 trans_color - graylevel or color sample values
3052 (in order red, green, blue) of the
3053 single transparent color for
3054 non-paletted images (PNG_INFO_tRNS)
3056 png_set_hIST(png_ptr, info_ptr, hist);
3058 hist - histogram of palette (array of
3059 png_uint_16) (PNG_INFO_hIST)
3061 png_set_tIME(png_ptr, info_ptr, mod_time);
3063 mod_time - time image was last modified
3066 png_set_bKGD(png_ptr, info_ptr, background);
3068 background - background color (of type
3069 png_color_16p) (PNG_VALID_bKGD)
3071 png_set_text(png_ptr, info_ptr, text_ptr, num_text);
3073 text_ptr - array of png_text holding image
3076 text_ptr[i].compression - type of compression used
3077 on "text" PNG_TEXT_COMPRESSION_NONE
3078 PNG_TEXT_COMPRESSION_zTXt
3079 PNG_ITXT_COMPRESSION_NONE
3080 PNG_ITXT_COMPRESSION_zTXt
3081 text_ptr[i].key - keyword for comment. Must contain
3083 text_ptr[i].text - text comments for current
3084 keyword. Can be NULL or empty.
3085 text_ptr[i].text_length - length of text string,
3086 after decompression, 0 for iTXt
3087 text_ptr[i].itxt_length - length of itxt string,
3088 after decompression, 0 for tEXt/zTXt
3089 text_ptr[i].lang - language of comment (NULL or
3091 text_ptr[i].translated_keyword - keyword in UTF-8 (NULL
3092 or empty for unknown).
3094 Note that the itxt_length, lang, and lang_key
3095 members of the text_ptr structure only exist when the
3096 library is built with iTXt chunk support. Prior to
3097 libpng-1.4.0 the library was built by default without
3098 iTXt support. Also note that when iTXt is supported,
3099 they contain NULL pointers when the "compression"
3100 field contains PNG_TEXT_COMPRESSION_NONE or
3101 PNG_TEXT_COMPRESSION_zTXt.
3103 num_text - number of comments
3105 png_set_sPLT(png_ptr, info_ptr, &palette_ptr,
3108 palette_ptr - array of png_sPLT_struct structures
3109 to be added to the list of palettes
3110 in the info structure.
3111 num_spalettes - number of palette structures to be
3114 png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y,
3117 offset_x - positive offset from the left
3120 offset_y - positive offset from the top
3123 unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
3125 png_set_pHYs(png_ptr, info_ptr, res_x, res_y,
3128 res_x - pixels/unit physical resolution
3131 res_y - pixels/unit physical resolution
3134 unit_type - PNG_RESOLUTION_UNKNOWN,
3135 PNG_RESOLUTION_METER
3137 png_set_sCAL(png_ptr, info_ptr, unit, width, height)
3139 unit - physical scale units (an integer)
3141 width - width of a pixel in physical scale units
3143 height - height of a pixel in physical scale units
3144 (width and height are doubles)
3146 png_set_sCAL_s(png_ptr, info_ptr, unit, width, height)
3148 unit - physical scale units (an integer)
3150 width - width of a pixel in physical scale units
3151 expressed as a string
3153 height - height of a pixel in physical scale units
3154 (width and height are strings like "2.54")
3156 png_set_unknown_chunks(png_ptr, info_ptr, &unknowns,
3159 unknowns - array of png_unknown_chunk
3160 structures holding unknown chunks
3161 unknowns[i].name - name of unknown chunk
3162 unknowns[i].data - data of unknown chunk
3163 unknowns[i].size - size of unknown chunk's data
3164 unknowns[i].location - position to write chunk in file
3165 0: do not write chunk
3166 PNG_HAVE_IHDR: before PLTE
3167 PNG_HAVE_PLTE: before IDAT
3168 PNG_AFTER_IDAT: after IDAT
3170 The "location" member is set automatically according to
3171 what part of the output file has already been written.
3172 You can change its value after calling png_set_unknown_chunks()
3173 as demonstrated in pngtest.c. Within each of the "locations",
3174 the chunks are sequenced according to their position in the
3175 structure (that is, the value of "i", which is the order in which
3176 the chunk was either read from the input file or defined with
3177 png_set_unknown_chunks).
3179 A quick word about text and num_text. text is an array of png_text
3180 structures. num_text is the number of valid structures in the array.
3181 Each png_text structure holds a language code, a keyword, a text value,
3182 and a compression type.
3184 The compression types have the same valid numbers as the compression
3185 types of the image data. Currently, the only valid number is zero.
3186 However, you can store text either compressed or uncompressed, unlike
3187 images, which always have to be compressed. So if you don't want the
3188 text compressed, set the compression type to PNG_TEXT_COMPRESSION_NONE.
3189 Because tEXt and zTXt chunks don't have a language field, if you
3190 specify PNG_TEXT_COMPRESSION_NONE or PNG_TEXT_COMPRESSION_zTXt
3191 any language code or translated keyword will not be written out.
3193 Until text gets around a few hundred bytes, it is not worth compressing it.
3194 After the text has been written out to the file, the compression type
3195 is set to PNG_TEXT_COMPRESSION_NONE_WR or PNG_TEXT_COMPRESSION_zTXt_WR,
3196 so that it isn't written out again at the end (in case you are calling
3197 png_write_end() with the same struct).
3199 The keywords that are given in the PNG Specification are:
3201 Title Short (one line) title or
3204 Author Name of image's creator
3206 Description Description of image (possibly long)
3208 Copyright Copyright notice
3210 Creation Time Time of original image creation
3211 (usually RFC 1123 format, see below)
3213 Software Software used to create the image
3215 Disclaimer Legal disclaimer
3217 Warning Warning of nature of content
3219 Source Device used to create the image
3221 Comment Miscellaneous comment; conversion
3222 from other image format
3224 The keyword-text pairs work like this. Keywords should be short
3225 simple descriptions of what the comment is about. Some typical
3226 keywords are found in the PNG specification, as is some recommendations
3227 on keywords. You can repeat keywords in a file. You can even write
3228 some text before the image and some after. For example, you may want
3229 to put a description of the image before the image, but leave the
3230 disclaimer until after, so viewers working over modem connections
3231 don't have to wait for the disclaimer to go over the modem before
3232 they start seeing the image. Finally, keywords should be full
3233 words, not abbreviations. Keywords and text are in the ISO 8859-1
3234 (Latin-1) character set (a superset of regular ASCII) and can not
3235 contain NUL characters, and should not contain control or other
3236 unprintable characters. To make the comments widely readable, stick
3237 with basic ASCII, and avoid machine specific character set extensions
3238 like the IBM-PC character set. The keyword must be present, but
3239 you can leave off the text string on non-compressed pairs.
3240 Compressed pairs must have a text string, as only the text string
3241 is compressed anyway, so the compression would be meaningless.
3243 PNG supports modification time via the png_time structure. Two
3244 conversion routines are provided, png_convert_from_time_t() for
3245 time_t and png_convert_from_struct_tm() for struct tm. The
3246 time_t routine uses gmtime(). You don't have to use either of
3247 these, but if you wish to fill in the png_time structure directly,
3248 you should provide the time in universal time (GMT) if possible
3249 instead of your local time. Note that the year number is the full
3250 year (e.g. 1998, rather than 98 - PNG is year 2000 compliant!), and
3251 that months start with 1.
3253 If you want to store the time of the original image creation, you should
3254 use a plain tEXt chunk with the "Creation Time" keyword. This is
3255 necessary because the "creation time" of a PNG image is somewhat vague,
3256 depending on whether you mean the PNG file, the time the image was
3257 created in a non-PNG format, a still photo from which the image was
3258 scanned, or possibly the subject matter itself. In order to facilitate
3259 machine-readable dates, it is recommended that the "Creation Time"
3260 tEXt chunk use RFC 1123 format dates (e.g. "22 May 1997 18:07:10 GMT"),
3261 although this isn't a requirement. Unlike the tIME chunk, the
3262 "Creation Time" tEXt chunk is not expected to be automatically changed
3263 by the software. To facilitate the use of RFC 1123 dates, a function
3264 png_convert_to_rfc1123_buffer(png_ptr, buffer, png_timep) is provided to
3265 convert from PNG time to an RFC 1123 format string. The caller must provide
3266 a writeable buffer of at least 29 bytes.
3268 Writing unknown chunks
3270 You can use the png_set_unknown_chunks function to queue up private chunks
3271 for writing. You give it a chunk name, location, raw data, and a size. You
3272 also must use png_set_keep_unknown_chunks() to ensure that libpng will
3273 handle them. That's all there is to it. The chunks will be written by the
3274 next following png_write_info_before_PLTE, png_write_info, or png_write_end
3275 function, depending upon the specified location. Any chunks previously
3276 read into the info structure's unknown-chunk list will also be written out
3277 in a sequence that satisfies the PNG specification's ordering rules.
3279 Here is an example of writing two private chunks, prVt and miNE:
3281 #ifdef PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED
3282 /* Set unknown chunk data */
3283 png_unknown_chunk unk_chunk[2];
3284 strcpy((char *) unk_chunk[0].name, "prVt";
3285 unk_chunk[0].data = (unsigned char *) "PRIVATE DATA";
3286 unk_chunk[0].size = strlen(unk_chunk[0].data)+1;
3287 unk_chunk[0].location = PNG_HAVE_IHDR;
3288 strcpy((char *) unk_chunk[1].name, "miNE";
3289 unk_chunk[1].data = (unsigned char *) "MY CHUNK DATA";
3290 unk_chunk[1].size = strlen(unk_chunk[0].data)+1;
3291 unk_chunk[1].location = PNG_AFTER_IDAT;
3292 png_set_unknown_chunks(write_ptr, write_info_ptr,
3294 /* Needed because miNE is not safe-to-copy */
3295 png_set_keep_unknown_chunks(png, PNG_HANDLE_CHUNK_ALWAYS,
3296 (png_bytep) "miNE", 1);
3297 # if PNG_LIBPNG_VER < 10600
3298 /* Deal with unknown chunk location bug in 1.5.x and earlier */
3299 png_set_unknown_chunk_location(png, info, 0, PNG_HAVE_IHDR);
3300 png_set_unknown_chunk_location(png, info, 1, PNG_AFTER_IDAT);
3302 # if PNG_LIBPNG_VER < 10500
3303 /* PNG_AFTER_IDAT writes two copies of the chunk prior to libpng-1.5.0,
3304 * one before IDAT and another after IDAT, so don't use it; only use
3305 * PNG_HAVE_IHDR location. This call resets the location previously
3306 * set by assignment and png_set_unknown_chunk_location() for chunk 1.
3308 png_set_unknown_chunk_location(png, info, 1, PNG_HAVE_IHDR);
3312 The high-level write interface
3314 At this point there are two ways to proceed; through the high-level
3315 write interface, or through a sequence of low-level write operations.
3316 You can use the high-level interface if your image data is present
3317 in the info structure. All defined output
3318 transformations are permitted, enabled by the following masks.
3320 PNG_TRANSFORM_IDENTITY No transformation
3321 PNG_TRANSFORM_PACKING Pack 1, 2 and 4-bit samples
3322 PNG_TRANSFORM_PACKSWAP Change order of packed
3324 PNG_TRANSFORM_INVERT_MONO Invert monochrome images
3325 PNG_TRANSFORM_SHIFT Normalize pixels to the
3327 PNG_TRANSFORM_BGR Flip RGB to BGR, RGBA
3329 PNG_TRANSFORM_SWAP_ALPHA Flip RGBA to ARGB or GA
3331 PNG_TRANSFORM_INVERT_ALPHA Change alpha from opacity
3333 PNG_TRANSFORM_SWAP_ENDIAN Byte-swap 16-bit samples
3334 PNG_TRANSFORM_STRIP_FILLER Strip out filler
3336 PNG_TRANSFORM_STRIP_FILLER_BEFORE Strip out leading
3338 PNG_TRANSFORM_STRIP_FILLER_AFTER Strip out trailing
3341 If you have valid image data in the info structure (you can use
3342 png_set_rows() to put image data in the info structure), simply do this:
3344 png_write_png(png_ptr, info_ptr, png_transforms, NULL)
3346 where png_transforms is an integer containing the bitwise OR of some set of
3347 transformation flags. This call is equivalent to png_write_info(),
3348 followed the set of transformations indicated by the transform mask,
3349 then png_write_image(), and finally png_write_end().
3351 (The final parameter of this call is not yet used. Someday it might point
3352 to transformation parameters required by some future output transform.)
3354 You must use png_transforms and not call any png_set_transform() functions
3355 when you use png_write_png().
3357 The low-level write interface
3359 If you are going the low-level route instead, you are now ready to
3360 write all the file information up to the actual image data. You do
3361 this with a call to png_write_info().
3363 png_write_info(png_ptr, info_ptr);
3365 Note that there is one transformation you may need to do before
3366 png_write_info(). In PNG files, the alpha channel in an image is the
3367 level of opacity. If your data is supplied as a level of transparency,
3368 you can invert the alpha channel before you write it, so that 0 is
3369 fully transparent and 255 (in 8-bit or paletted images) or 65535
3370 (in 16-bit images) is fully opaque, with
3372 png_set_invert_alpha(png_ptr);
3374 This must appear before png_write_info() instead of later with the
3375 other transformations because in the case of paletted images the tRNS
3376 chunk data has to be inverted before the tRNS chunk is written. If
3377 your image is not a paletted image, the tRNS data (which in such cases
3378 represents a single color to be rendered as transparent) won't need to
3379 be changed, and you can safely do this transformation after your
3380 png_write_info() call.
3382 If you need to write a private chunk that you want to appear before
3383 the PLTE chunk when PLTE is present, you can write the PNG info in
3384 two steps, and insert code to write your own chunk between them:
3386 png_write_info_before_PLTE(png_ptr, info_ptr);
3387 png_set_unknown_chunks(png_ptr, info_ptr, ...);
3388 png_write_info(png_ptr, info_ptr);
3390 After you've written the file information, you can set up the library
3391 to handle any special transformations of the image data. The various
3392 ways to transform the data will be described in the order that they
3393 should occur. This is important, as some of these change the color
3394 type and/or bit depth of the data, and some others only work on
3395 certain color types and bit depths. Even though each transformation
3396 checks to see if it has data that it can do something with, you should
3397 make sure to only enable a transformation if it will be valid for the
3398 data. For example, don't swap red and blue on grayscale data.
3400 PNG files store RGB pixels packed into 3 or 6 bytes. This code tells
3401 the library to strip input data that has 4 or 8 bytes per pixel down
3402 to 3 or 6 bytes (or strip 2 or 4-byte grayscale+filler data to 1 or 2
3405 png_set_filler(png_ptr, 0, PNG_FILLER_BEFORE);
3407 where the 0 is unused, and the location is either PNG_FILLER_BEFORE or
3408 PNG_FILLER_AFTER, depending upon whether the filler byte in the pixel
3409 is stored XRGB or RGBX.
3411 PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
3412 they can, resulting in, for example, 8 pixels per byte for 1 bit files.
3413 If the data is supplied at 1 pixel per byte, use this code, which will
3414 correctly pack the pixels into a single byte:
3416 png_set_packing(png_ptr);
3418 PNG files reduce possible bit depths to 1, 2, 4, 8, and 16. If your
3419 data is of another bit depth, you can write an sBIT chunk into the
3420 file so that decoders can recover the original data if desired.
3422 /* Set the true bit depth of the image data */
3423 if (color_type & PNG_COLOR_MASK_COLOR)
3425 sig_bit.red = true_bit_depth;
3426 sig_bit.green = true_bit_depth;
3427 sig_bit.blue = true_bit_depth;
3432 sig_bit.gray = true_bit_depth;
3435 if (color_type & PNG_COLOR_MASK_ALPHA)
3437 sig_bit.alpha = true_bit_depth;
3440 png_set_sBIT(png_ptr, info_ptr, &sig_bit);
3442 If the data is stored in the row buffer in a bit depth other than
3443 one supported by PNG (e.g. 3 bit data in the range 0-7 for a 4-bit PNG),
3444 this will scale the values to appear to be the correct bit depth as
3447 png_set_shift(png_ptr, &sig_bit);
3449 PNG files store 16-bit pixels in network byte order (big-endian,
3450 ie. most significant bits first). This code would be used if they are
3451 supplied the other way (little-endian, i.e. least significant bits
3452 first, the way PCs store them):
3455 png_set_swap(png_ptr);
3457 If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
3458 need to change the order the pixels are packed into bytes, you can use:
3461 png_set_packswap(png_ptr);
3463 PNG files store 3 color pixels in red, green, blue order. This code
3464 would be used if they are supplied as blue, green, red:
3466 png_set_bgr(png_ptr);
3468 PNG files describe monochrome as black being zero and white being
3469 one. This code would be used if the pixels are supplied with this reversed
3470 (black being one and white being zero):
3472 png_set_invert_mono(png_ptr);
3474 Finally, you can write your own transformation function if none of
3475 the existing ones meets your needs. This is done by setting a callback
3478 png_set_write_user_transform_fn(png_ptr,
3479 write_transform_fn);
3481 You must supply the function
3483 void write_transform_fn(png_structp png_ptr, png_row_infop
3484 row_info, png_bytep data)
3486 See pngtest.c for a working example. Your function will be called
3487 before any of the other transformations are processed. If supported
3488 libpng also supplies an information routine that may be called from
3491 png_get_current_row_number(png_ptr);
3492 png_get_current_pass_number(png_ptr);
3494 This returns the current row passed to the transform. With interlaced
3495 images the value returned is the row in the input sub-image image. Use
3496 PNG_ROW_FROM_PASS_ROW(row, pass) and PNG_COL_FROM_PASS_COL(col, pass) to
3497 find the output pixel (x,y) given an interlaced sub-image pixel (row,col,pass).
3499 The discussion of interlace handling above contains more information on how to
3502 You can also set up a pointer to a user structure for use by your
3505 png_set_user_transform_info(png_ptr, user_ptr, 0, 0);
3507 The user_channels and user_depth parameters of this function are ignored
3508 when writing; you can set them to zero as shown.
3510 You can retrieve the pointer via the function png_get_user_transform_ptr().
3513 voidp write_user_transform_ptr =
3514 png_get_user_transform_ptr(png_ptr);
3516 It is possible to have libpng flush any pending output, either manually,
3517 or automatically after a certain number of lines have been written. To
3518 flush the output stream a single time call:
3520 png_write_flush(png_ptr);
3522 and to have libpng flush the output stream periodically after a certain
3523 number of scanlines have been written, call:
3525 png_set_flush(png_ptr, nrows);
3527 Note that the distance between rows is from the last time png_write_flush()
3528 was called, or the first row of the image if it has never been called.
3529 So if you write 50 lines, and then png_set_flush 25, it will flush the
3530 output on the next scanline, and every 25 lines thereafter, unless
3531 png_write_flush() is called before 25 more lines have been written.
3532 If nrows is too small (less than about 10 lines for a 640 pixel wide
3533 RGB image) the image compression may decrease noticeably (although this
3534 may be acceptable for real-time applications). Infrequent flushing will
3535 only degrade the compression performance by a few percent over images
3536 that do not use flushing.
3538 Writing the image data
3540 That's it for the transformations. Now you can write the image data.
3541 The simplest way to do this is in one function call. If you have the
3542 whole image in memory, you can just call png_write_image() and libpng
3543 will write the image. You will need to pass in an array of pointers to
3544 each row. This function automatically handles interlacing, so you don't
3545 need to call png_set_interlace_handling() or call this function multiple
3546 times, or any of that other stuff necessary with png_write_rows().
3548 png_write_image(png_ptr, row_pointers);
3550 where row_pointers is:
3552 png_byte *row_pointers[height];
3554 You can point to void or char or whatever you use for pixels.
3556 If you don't want to write the whole image at once, you can
3557 use png_write_rows() instead. If the file is not interlaced,
3560 png_write_rows(png_ptr, row_pointers,
3563 row_pointers is the same as in the png_write_image() call.
3565 If you are just writing one row at a time, you can do this with
3566 a single row_pointer instead of an array of row_pointers:
3568 png_bytep row_pointer = row;
3570 png_write_row(png_ptr, row_pointer);
3572 When the file is interlaced, things can get a good deal more complicated.
3573 The only currently (as of the PNG Specification version 1.2, dated July
3574 1999) defined interlacing scheme for PNG files is the "Adam7" interlace
3575 scheme, that breaks down an image into seven smaller images of varying
3576 size. libpng will build these images for you, or you can do them
3577 yourself. If you want to build them yourself, see the PNG specification
3578 for details of which pixels to write when.
3580 If you don't want libpng to handle the interlacing details, just
3581 use png_set_interlace_handling() and call png_write_rows() the
3582 correct number of times to write all the sub-images
3583 (png_set_interlace_handling() returns the number of sub-images.)
3585 If you want libpng to build the sub-images, call this before you start
3588 number_of_passes = png_set_interlace_handling(png_ptr);
3590 This will return the number of passes needed. Currently, this is seven,
3591 but may change if another interlace type is added.
3593 Then write the complete image number_of_passes times.
3595 png_write_rows(png_ptr, row_pointers, number_of_rows);
3597 Think carefully before you write an interlaced image. Typically code that
3598 reads such images reads all the image data into memory, uncompressed, before
3599 doing any processing. Only code that can display an image on the fly can
3600 take advantage of the interlacing and even then the image has to be exactly
3601 the correct size for the output device, because scaling an image requires
3602 adjacent pixels and these are not available until all the passes have been
3605 If you do write an interlaced image you will hardly ever need to handle
3606 the interlacing yourself. Call png_set_interlace_handling() and use the
3607 approach described above.
3609 The only time it is conceivable that you will really need to write an
3610 interlaced image pass-by-pass is when you have read one pass by pass and
3611 made some pixel-by-pixel transformation to it, as described in the read
3612 code above. In this case use the PNG_PASS_ROWS and PNG_PASS_COLS macros
3613 to determine the size of each sub-image in turn and simply write the rows
3614 you obtained from the read code.
3616 Finishing a sequential write
3618 After you are finished writing the image, you should finish writing
3619 the file. If you are interested in writing comments or time, you should
3620 pass an appropriately filled png_info pointer. If you are not interested,
3623 png_write_end(png_ptr, info_ptr);
3625 When you are done, you can free all memory used by libpng like this:
3627 png_destroy_write_struct(&png_ptr, &info_ptr);
3629 It is also possible to individually free the info_ptr members that
3630 point to libpng-allocated storage with the following function:
3632 png_free_data(png_ptr, info_ptr, mask, seq)
3634 mask - identifies data to be freed, a mask
3635 containing the bitwise OR of one or
3637 PNG_FREE_PLTE, PNG_FREE_TRNS,
3638 PNG_FREE_HIST, PNG_FREE_ICCP,
3639 PNG_FREE_PCAL, PNG_FREE_ROWS,
3640 PNG_FREE_SCAL, PNG_FREE_SPLT,
3641 PNG_FREE_TEXT, PNG_FREE_UNKN,
3642 or simply PNG_FREE_ALL
3644 seq - sequence number of item to be freed
3647 This function may be safely called when the relevant storage has
3648 already been freed, or has not yet been allocated, or was allocated
3649 by the user and not by libpng, and will in those cases do nothing.
3650 The "seq" parameter is ignored if only one item of the selected data
3651 type, such as PLTE, is allowed. If "seq" is not -1, and multiple items
3652 are allowed for the data type identified in the mask, such as text or
3653 sPLT, only the n'th item in the structure is freed, where n is "seq".
3655 If you allocated data such as a palette that you passed in to libpng
3656 with png_set_*, you must not free it until just before the call to
3657 png_destroy_write_struct().
3659 The default behavior is only to free data that was allocated internally
3660 by libpng. This can be changed, so that libpng will not free the data,
3661 or so that it will free data that was allocated by the user with png_malloc()
3662 or png_calloc() and passed in via a png_set_*() function, with
3664 png_data_freer(png_ptr, info_ptr, freer, mask)
3667 PNG_DESTROY_WILL_FREE_DATA
3668 PNG_SET_WILL_FREE_DATA
3669 PNG_USER_WILL_FREE_DATA
3671 mask - which data elements are affected
3672 same choices as in png_free_data()
3674 For example, to transfer responsibility for some data from a read structure
3675 to a write structure, you could use
3677 png_data_freer(read_ptr, read_info_ptr,
3678 PNG_USER_WILL_FREE_DATA,
3679 PNG_FREE_PLTE|PNG_FREE_tRNS|PNG_FREE_hIST)
3681 png_data_freer(write_ptr, write_info_ptr,
3682 PNG_DESTROY_WILL_FREE_DATA,
3683 PNG_FREE_PLTE|PNG_FREE_tRNS|PNG_FREE_hIST)
3685 thereby briefly reassigning responsibility for freeing to the user but
3686 immediately afterwards reassigning it once more to the write_destroy
3687 function. Having done this, it would then be safe to destroy the read
3688 structure and continue to use the PLTE, tRNS, and hIST data in the write
3691 This function only affects data that has already been allocated.
3692 You can call this function before calling after the png_set_*() functions
3693 to control whether the user or png_destroy_*() is supposed to free the data.
3694 When the user assumes responsibility for libpng-allocated data, the
3695 application must use
3696 png_free() to free it, and when the user transfers responsibility to libpng
3697 for data that the user has allocated, the user must have used png_malloc()
3698 or png_calloc() to allocate it.
3700 If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
3701 separately, do not transfer responsibility for freeing text_ptr to libpng,
3702 because when libpng fills a png_text structure it combines these members with
3703 the key member, and png_free_data() will free only text_ptr.key. Similarly,
3704 if you transfer responsibility for free'ing text_ptr from libpng to your
3705 application, your application must not separately free those members.
3706 For a more compact example of writing a PNG image, see the file example.c.
3710 The simplified API, which became available in libpng-1.6.0, hides the details
3711 of both libpng and the PNG file format itself.
3712 It allows PNG files to be read into a very limited number of
3713 in-memory bitmap formats or to be written from the same formats. If these
3714 formats do not accommodate your needs then you can, and should, use the more
3715 sophisticated APIs above - these support a wide variety of in-memory formats
3716 and a wide variety of sophisticated transformations to those formats as well
3717 as a wide variety of APIs to manipulate ancilliary information.
3719 To read a PNG file using the simplified API:
3721 1) Declare a 'png_image' structure (see below) on the
3722 stack and memset() it to all zero.
3724 2) Call the appropriate png_image_begin_read... function.
3726 3) Set the png_image 'format' member to the required
3727 format and allocate a buffer for the image.
3729 4) Call png_image_finish_read to read the image into
3732 There are no restrictions on the format of the PNG input itself; all valid
3733 color types, bit depths, and interlace methods are acceptable, and the
3734 input image is transformed as necessary to the requested in-memory format
3735 during the png_image_finish_read() step.
3737 To write a PNG file using the simplified API:
3739 1) Declare a 'png_image' structure on the stack and memset()
3742 2) Initialize the members of the structure that describe the
3743 image, setting the 'format' member to the format of the
3746 3) Call the appropriate png_image_write... function with a
3747 pointer to the image to write the PNG data.
3749 png_image is a structure that describes the in-memory format of an image
3750 when it is being read or define the in-memory format of an image that you
3751 need to write. The "png_image" structure contains the following members:
3753 png_uint_32 version Set to PNG_IMAGE_VERSION
3754 png_uint_32 width Image width in pixels (columns)
3755 png_uint_32 height Image height in pixels (rows)
3756 png_uint_32 format Image format as defined below
3757 png_uint_32 flags A bit mask containing informational flags
3758 png_controlp opaque Initialize to NULL, free with png_image_free
3759 png_uint_32 colormap_entries; Number of entries in the color-map
3760 png_uint_32 warning_or_error;
3763 In the event of an error or warning the following field warning_or_error
3764 field will be set to a non-zero value and the 'message' field will contain
3765 a '\0' terminated string with the libpng error or warning message. If both
3766 warnings and an error were encountered, only the error is recorded. If there
3767 are multiple warnings, only the first one is recorded.
3769 The upper 30 bits of this value are reserved; the low two bits contain
3770 a two bit code such that a value more than 1 indicates a failure in the API
3773 0 - no warning or error
3776 3 - error preceded by warning
3778 The pixels (samples) of the image have one to four channels whose components
3779 have original values in the range 0 to 1.0:
3781 1: A single gray or luminance channel (G).
3782 2: A gray/luminance channel and an alpha channel (GA).
3783 3: Three red, green, blue color channels (RGB).
3784 4: Three color channels and an alpha channel (RGBA).
3786 The channels are encoded in one of two ways:
3788 a) As a small integer, value 0..255, contained in a single byte. For the
3789 alpha channel the original value is simply value/255. For the color or
3790 luminance channels the value is encoded according to the sRGB specification
3791 and matches the 8-bit format expected by typical display devices.
3793 The color/gray channels are not scaled (pre-multiplied) by the alpha
3794 channel and are suitable for passing to color management software.
3796 b) As a value in the range 0..65535, contained in a 2-byte integer, in
3797 the native byte order of the platform on which the application is running.
3798 All channels can be converted to the original value by dividing by 65535; all
3799 channels are linear. Color channels use the RGB encoding (RGB end-points) of
3800 the sRGB specification. This encoding is identified by the
3801 PNG_FORMAT_FLAG_LINEAR flag below.
3803 When an alpha channel is present it is expected to denote pixel coverage
3804 of the color or luminance channels and is returned as an associated alpha
3805 channel: the color/gray channels are scaled (pre-multiplied) by the alpha
3808 When a color-mapped image is used as a result of calling
3809 png_image_read_colormap or png_image_write_colormap the channels are encoded
3810 in the color-map and the descriptions above apply to the color-map entries.
3811 The image data is encoded as small integers, value 0..255, that index the
3812 entries in the color-map. One integer (one byte) is stored for each pixel.
3816 The #defines to be used in png_image::format. Each #define identifies a
3817 particular layout of channel data and, if present, alpha values. There are
3818 separate defines for each of the two channel encodings.
3820 A format is built up using single bit flag values. Not all combinations are
3821 valid: use the bit flag values below for testing a format returned by the
3822 read APIs, but set formats from the derived values.
3824 When reading or writing color-mapped images the format should be set to the
3825 format of the entries in the color-map then png_image_{read,write}_colormap
3826 called to read or write the color-map and set the format correctly for the
3827 image data. Do not set the PNG_FORMAT_FLAG_COLORMAP bit directly!
3829 NOTE: libpng can be built with particular features disabled, if you see
3830 compiler errors because the definition of one of the following flags has been
3831 compiled out it is because libpng does not have the required support. It is
3832 possible, however, for the libpng configuration to enable the format on just
3833 read or just write; in that case you may see an error at run time. You can
3834 guard against this by checking for the definition of:
3836 PNG_SIMPLIFIED_{READ,WRITE}_{BGR,AFIRST}_SUPPORTED
3838 PNG_FORMAT_FLAG_ALPHA 0x01 format with an alpha channel
3839 PNG_FORMAT_FLAG_COLOR 0x02 color format: otherwise grayscale
3840 PNG_FORMAT_FLAG_LINEAR 0x04 png_uint_16 channels else png_byte
3841 PNG_FORMAT_FLAG_COLORMAP 0x08 libpng use only
3842 PNG_FORMAT_FLAG_BGR 0x10 BGR colors, else order is RGB
3843 PNG_FORMAT_FLAG_AFIRST 0x20 alpha channel comes first
3845 Supported formats are as follows. Future versions of libpng may support more
3846 formats; for compatibility with older versions simply check if the format
3847 macro is defined using #ifdef. These defines describe the in-memory layout
3848 of the components of the pixels of the image.
3850 First the single byte formats:
3853 PNG_FORMAT_GA PNG_FORMAT_FLAG_ALPHA
3854 PNG_FORMAT_AG (PNG_FORMAT_GA|PNG_FORMAT_FLAG_AFIRST)
3855 PNG_FORMAT_RGB PNG_FORMAT_FLAG_COLOR
3856 PNG_FORMAT_BGR (PNG_FORMAT_FLAG_COLOR|PNG_FORMAT_FLAG_BGR)
3857 PNG_FORMAT_RGBA (PNG_FORMAT_RGB|PNG_FORMAT_FLAG_ALPHA)
3858 PNG_FORMAT_ARGB (PNG_FORMAT_RGBA|PNG_FORMAT_FLAG_AFIRST)
3859 PNG_FORMAT_BGRA (PNG_FORMAT_BGR|PNG_FORMAT_FLAG_ALPHA)
3860 PNG_FORMAT_ABGR (PNG_FORMAT_BGRA|PNG_FORMAT_FLAG_AFIRST)
3862 Then the linear 2-byte formats. When naming these "Y" is used to
3863 indicate a luminance (gray) channel. The component order within the pixel
3864 is always the same - there is no provision for swapping the order of the
3865 components in the linear format. The components are 16-bit integers in
3866 the native byte order for your platform, and there is no provision for
3867 swapping the bytes to a different endian condition.
3869 PNG_FORMAT_LINEAR_Y PNG_FORMAT_FLAG_LINEAR
3870 PNG_FORMAT_LINEAR_Y_ALPHA
3871 (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_ALPHA)
3872 PNG_FORMAT_LINEAR_RGB
3873 (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_COLOR)
3874 PNG_FORMAT_LINEAR_RGB_ALPHA
3875 (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_COLOR|
3876 PNG_FORMAT_FLAG_ALPHA)
3878 Color-mapped formats are obtained by calling png_image_{read,write}_colormap,
3879 as appropriate after setting png_image::format to the format of the color-map
3880 to be read or written. Applications may check the value of
3881 PNG_FORMAT_FLAG_COLORMAP to see if they have called the colormap API. The
3882 format of the color-map may be extracted using the following macro.
3884 PNG_FORMAT_OF_COLORMAP(fmt) ((fmt) & ~PNG_FORMAT_FLAG_COLORMAP)
3888 These are convenience macros to derive information from a png_image
3889 structure. The PNG_IMAGE_SAMPLE_ macros return values appropriate to the
3890 actual image sample values - either the entries in the color-map or the
3891 pixels in the image. The PNG_IMAGE_PIXEL_ macros return corresponding values
3892 for the pixels and will always return 1 after a call to
3893 png_image_{read,write}_colormap. The remaining macros return information
3894 about the rows in the image and the complete image.
3896 NOTE: All the macros that take a png_image::format parameter are compile time
3897 constants if the format parameter is, itself, a constant. Therefore these
3898 macros can be used in array declarations and case labels where required.
3899 Similarly the macros are also pre-processor constants (sizeof is not used) so
3900 they can be used in #if tests.
3902 First the information about the samples.
3904 PNG_IMAGE_SAMPLE_CHANNELS(fmt)
3905 Returns the total number of channels in a given format: 1..4
3907 PNG_IMAGE_SAMPLE_COMPONENT_SIZE(fmt)
3908 Returns the size in bytes of a single component of a pixel or color-map
3909 entry (as appropriate) in the image.
3911 PNG_IMAGE_SAMPLE_SIZE(fmt)
3912 This is the size of the sample data for one sample. If the image is
3913 color-mapped it is the size of one color-map entry (and image pixels are
3914 one byte in size), otherwise it is the size of one image pixel.
3916 PNG_IMAGE_COLORMAP_SIZE(fmt)
3917 The size of the color-map required by the format; this is the size of the
3918 color-map buffer passed to the png_image_{read,write}_colormap APIs, it is
3919 a fixed number determined by the format so can easily be allocated on the
3922 #define PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(fmt)\
3923 (PNG_IMAGE_SAMPLE_CHANNELS(fmt) * 256)
3924 /* The maximum size of the color-map required by the format expressed in a
3925 * count of components. This can be used to compile-time allocate a
3928 * png_uint_16 colormap[PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(linear_fmt)];
3930 * png_byte colormap[PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(sRGB_fmt)];
3932 * Alternatively, use the PNG_IMAGE_COLORMAP_SIZE macro below to use the
3933 * information from one of the png_image_begin_read_ APIs and dynamically
3934 * allocate the required memory.
3938 Corresponding information about the pixels
3940 PNG_IMAGE_PIXEL_(test,fmt)
3942 PNG_IMAGE_PIXEL_CHANNELS(fmt)
3943 The number of separate channels (components) in a pixel; 1 for a
3946 PNG_IMAGE_PIXEL_COMPONENT_SIZE(fmt)\
3947 The size, in bytes, of each component in a pixel; 1 for a color-mapped
3950 PNG_IMAGE_PIXEL_SIZE(fmt)
3951 The size, in bytes, of a complete pixel; 1 for a color-mapped image.
3953 Information about the whole row, or whole image
3955 PNG_IMAGE_ROW_STRIDE(image)
3956 Returns the total number of components in a single row of the image; this
3957 is the minimum 'row stride', the minimum count of components between each
3958 row. For a color-mapped image this is the minimum number of bytes in a
3961 If you need the stride measured in bytes, row_stride_bytes is
3962 PNG_IMAGE_ROW_STRIDE(image) * PNG_IMAGE_PIXEL_COMPONENT_SIZE(fmt)
3963 plus any padding bytes that your application might need, for example
3964 to start the next row on a 4-byte boundary.
3966 PNG_IMAGE_BUFFER_SIZE(image, row_stride)
3967 Returns the size, in bytes, of an image buffer given a png_image and a row
3968 stride - the number of components to leave space for in each row. This
3969 macro takes care of multiplying row_stride by PNG_IMAGE_PIXEL_COMONENT_SIZE
3970 when the image has 2-byte components.
3972 PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB == 0x01
3973 This indicates the the RGB values of the in-memory bitmap do not
3974 correspond to the red, green and blue end-points defined by sRGB.
3976 PNG_IMAGE_FLAG_COLORMAP == 0x02
3977 The PNG is color-mapped. If this flag is set png_image_read_colormap
3978 can be used without further loss of image information. If it is not set
3979 png_image_read_colormap will cause significant loss if the image has any
3983 The png_image passed to the read APIs must have been initialized by setting
3984 the png_controlp field 'opaque' to NULL (or, better, memset the whole thing.)
3986 int png_image_begin_read_from_file( png_imagep image,
3987 const char *file_name)
3989 The named file is opened for read and the image header
3990 is filled in from the PNG header in the file.
3992 int png_image_begin_read_from_stdio (png_imagep image,
3995 The PNG header is read from the stdio FILE object.
3997 int png_image_begin_read_from_memory(png_imagep image,
3998 png_const_voidp memory, png_size_t size)
4000 The PNG header is read from the given memory buffer.
4002 int png_image_finish_read(png_imagep image,
4003 png_colorp background, void *buffer,
4004 png_int_32 row_stride, void *colormap));
4006 Finish reading the image into the supplied buffer and
4007 clean up the png_image structure.
4009 row_stride is the step, in png_byte or png_uint_16 units
4010 as appropriate, between adjacent rows. A positive stride
4011 indicates that the top-most row is first in the buffer -
4012 the normal top-down arrangement. A negative stride
4013 indicates that the bottom-most row is first in the buffer.
4015 background need only be supplied if an alpha channel must
4016 be removed from a png_byte format and the removal is to be
4017 done by compositing on a solid color; otherwise it may be
4018 NULL and any composition will be done directly onto the
4019 buffer. The value is an sRGB color to use for the
4020 background, for grayscale output the green channel is used.
4022 For linear output removing the alpha channel is always done
4023 by compositing on black.
4025 void png_image_free(png_imagep image)
4027 Free any data allocated by libpng in image->opaque,
4028 setting the pointer to NULL. May be called at any time
4029 after the structure is initialized.
4031 When the simplified API needs to convert between sRGB and linear colorspaces,
4032 the actual sRGB transfer curve defined in the sRGB specification (see the
4033 article at http://en.wikipedia.org/wiki/SRGB) is used, not the gamma=1/2.2
4034 approximation used elsewhere in libpng.
4038 For write you must initialize a png_image structure to describe the image to
4041 version: must be set to PNG_IMAGE_VERSION
4042 opaque: must be initialized to NULL
4043 width: image width in pixels
4044 height: image height in rows
4045 format: the format of the data you wish to write
4046 flags: set to 0 unless one of the defined flags applies; set
4047 PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB for color format images
4048 where the RGB values do not correspond to the colors in sRGB.
4049 colormap_entries: set to the number of entries in the color-map (0 to 256)
4051 int png_image_write_to_file, (png_imagep image,
4052 const char *file, int convert_to_8bit, const void *buffer,
4053 png_int_32 row_stride, const void *colormap));
4055 Write the image to the named file.
4057 int png_image_write_to_stdio(png_imagep image, FILE *file,
4058 int convert_to_8_bit, const void *buffer,
4059 png_int_32 row_stride, const void *colormap)
4061 Write the image to the given (FILE*).
4063 With all write APIs if image is in one of the linear formats with
4064 (png_uint_16) data then setting convert_to_8_bit will cause the output to be
4065 a (png_byte) PNG gamma encoded according to the sRGB specification, otherwise
4066 a 16-bit linear encoded PNG file is written.
4068 With all APIs row_stride is handled as in the read APIs - it is the spacing
4069 from one row to the next in component sized units (float) and if negative
4070 indicates a bottom-up row layout in the buffer.
4072 Note that the write API does not support interlacing, sub-8-bit pixels,
4073 and indexed (paletted) images.
4075 VI. Modifying/Customizing libpng
4077 There are two issues here. The first is changing how libpng does
4078 standard things like memory allocation, input/output, and error handling.
4079 The second deals with more complicated things like adding new chunks,
4080 adding new transformations, and generally changing how libpng works.
4081 Both of those are compile-time issues; that is, they are generally
4082 determined at the time the code is written, and there is rarely a need
4083 to provide the user with a means of changing them.
4085 Memory allocation, input/output, and error handling
4087 All of the memory allocation, input/output, and error handling in libpng
4088 goes through callbacks that are user-settable. The default routines are
4089 in pngmem.c, pngrio.c, pngwio.c, and pngerror.c, respectively. To change
4090 these functions, call the appropriate png_set_*_fn() function.
4092 Memory allocation is done through the functions png_malloc(), png_calloc(),
4093 and png_free(). The png_malloc() and png_free() functions currently just
4094 call the standard C functions and png_calloc() calls png_malloc() and then
4095 clears the newly allocated memory to zero; note that png_calloc(png_ptr, size)
4096 is not the same as the calloc(number, size) function provided by stdlib.h.
4097 There is limited support for certain systems with segmented memory
4098 architectures and the types of pointers declared by png.h match this; you
4099 will have to use appropriate pointers in your application. If you prefer
4100 to use a different method of allocating and freeing data, you can use
4101 png_create_read_struct_2() or png_create_write_struct_2() to register your
4102 own functions as described above. These functions also provide a void
4103 pointer that can be retrieved via
4105 mem_ptr=png_get_mem_ptr(png_ptr);
4107 Your replacement memory functions must have prototypes as follows:
4109 png_voidp malloc_fn(png_structp png_ptr,
4110 png_alloc_size_t size);
4112 void free_fn(png_structp png_ptr, png_voidp ptr);
4114 Your malloc_fn() must return NULL in case of failure. The png_malloc()
4115 function will normally call png_error() if it receives a NULL from the
4116 system memory allocator or from your replacement malloc_fn().
4118 Your free_fn() will never be called with a NULL ptr, since libpng's
4119 png_free() checks for NULL before calling free_fn().
4121 Input/Output in libpng is done through png_read() and png_write(),
4122 which currently just call fread() and fwrite(). The FILE * is stored in
4123 png_struct and is initialized via png_init_io(). If you wish to change
4124 the method of I/O, the library supplies callbacks that you can set
4125 through the function png_set_read_fn() and png_set_write_fn() at run
4126 time, instead of calling the png_init_io() function. These functions
4127 also provide a void pointer that can be retrieved via the function
4128 png_get_io_ptr(). For example:
4130 png_set_read_fn(png_structp read_ptr,
4131 voidp read_io_ptr, png_rw_ptr read_data_fn)
4133 png_set_write_fn(png_structp write_ptr,
4134 voidp write_io_ptr, png_rw_ptr write_data_fn,
4135 png_flush_ptr output_flush_fn);
4137 voidp read_io_ptr = png_get_io_ptr(read_ptr);
4138 voidp write_io_ptr = png_get_io_ptr(write_ptr);
4140 The replacement I/O functions must have prototypes as follows:
4142 void user_read_data(png_structp png_ptr,
4143 png_bytep data, png_size_t length);
4145 void user_write_data(png_structp png_ptr,
4146 png_bytep data, png_size_t length);
4148 void user_flush_data(png_structp png_ptr);
4150 The user_read_data() function is responsible for detecting and
4151 handling end-of-data errors.
4153 Supplying NULL for the read, write, or flush functions sets them back
4154 to using the default C stream functions, which expect the io_ptr to
4155 point to a standard *FILE structure. It is probably a mistake
4156 to use NULL for one of write_data_fn and output_flush_fn but not both
4157 of them, unless you have built libpng with PNG_NO_WRITE_FLUSH defined.
4158 It is an error to read from a write stream, and vice versa.
4160 Error handling in libpng is done through png_error() and png_warning().
4161 Errors handled through png_error() are fatal, meaning that png_error()
4162 should never return to its caller. Currently, this is handled via
4163 setjmp() and longjmp() (unless you have compiled libpng with
4164 PNG_NO_SETJMP, in which case it is handled via PNG_ABORT()),
4165 but you could change this to do things like exit() if you should wish,
4166 as long as your function does not return.
4168 On non-fatal errors, png_warning() is called
4169 to print a warning message, and then control returns to the calling code.
4170 By default png_error() and png_warning() print a message on stderr via
4171 fprintf() unless the library is compiled with PNG_NO_CONSOLE_IO defined
4172 (because you don't want the messages) or PNG_NO_STDIO defined (because
4173 fprintf() isn't available). If you wish to change the behavior of the error
4174 functions, you will need to set up your own message callbacks. These
4175 functions are normally supplied at the time that the png_struct is created.
4176 It is also possible to redirect errors and warnings to your own replacement
4177 functions after png_create_*_struct() has been called by calling:
4179 png_set_error_fn(png_structp png_ptr,
4180 png_voidp error_ptr, png_error_ptr error_fn,
4181 png_error_ptr warning_fn);
4183 png_voidp error_ptr = png_get_error_ptr(png_ptr);
4185 If NULL is supplied for either error_fn or warning_fn, then the libpng
4186 default function will be used, calling fprintf() and/or longjmp() if a
4187 problem is encountered. The replacement error functions should have
4188 parameters as follows:
4190 void user_error_fn(png_structp png_ptr,
4191 png_const_charp error_msg);
4193 void user_warning_fn(png_structp png_ptr,
4194 png_const_charp warning_msg);
4196 The motivation behind using setjmp() and longjmp() is the C++ throw and
4197 catch exception handling methods. This makes the code much easier to write,
4198 as there is no need to check every return code of every function call.
4199 However, there are some uncertainties about the status of local variables
4200 after a longjmp, so the user may want to be careful about doing anything
4201 after setjmp returns non-zero besides returning itself. Consult your
4202 compiler documentation for more details. For an alternative approach, you
4203 may wish to use the "cexcept" facility (see http://cexcept.sourceforge.net),
4204 which is illustrated in pngvalid.c and in contrib/visupng.
4206 Beginning in libpng-1.4.0, the png_set_benign_errors() API became available.
4207 You can use this to handle certain errors (normally handled as errors)
4210 png_set_benign_errors (png_ptr, int allowed);
4212 allowed: 0: treat png_benign_error() as an error.
4213 1: treat png_benign_error() as a warning.
4215 As of libpng-1.6.0, the default condition is to treat benign errors as
4216 warnings while reading and as errors while writing.
4220 If you need to read or write custom chunks, you may need to get deeper
4221 into the libpng code. The library now has mechanisms for storing
4222 and writing chunks of unknown type; you can even declare callbacks
4223 for custom chunks. However, this may not be good enough if the
4224 library code itself needs to know about interactions between your
4225 chunk and existing `intrinsic' chunks.
4227 If you need to write a new intrinsic chunk, first read the PNG
4228 specification. Acquire a first level of understanding of how it works.
4229 Pay particular attention to the sections that describe chunk names,
4230 and look at how other chunks were designed, so you can do things
4231 similarly. Second, check out the sections of libpng that read and
4232 write chunks. Try to find a chunk that is similar to yours and use
4233 it as a template. More details can be found in the comments inside
4234 the code. It is best to handle private or unknown chunks in a generic method,
4235 via callback functions, instead of by modifying libpng functions. This
4236 is illustrated in pngtest.c, which uses a callback function to handle a
4237 private "vpAg" chunk and the new "sTER" chunk, which are both unknown to
4240 If you wish to write your own transformation for the data, look through
4241 the part of the code that does the transformations, and check out some of
4242 the simpler ones to get an idea of how they work. Try to find a similar
4243 transformation to the one you want to add and copy off of it. More details
4244 can be found in the comments inside the code itself.
4246 Configuring for gui/windowing platforms:
4248 You will need to write new error and warning functions that use the GUI
4249 interface, as described previously, and set them to be the error and
4250 warning functions at the time that png_create_*_struct() is called,
4251 in order to have them available during the structure initialization.
4252 They can be changed later via png_set_error_fn(). On some compilers,
4253 you may also have to change the memory allocators (png_malloc, etc.).
4257 There are special functions to configure the compression. Perhaps the
4258 most useful one changes the compression level, which currently uses
4259 input compression values in the range 0 - 9. The library normally
4260 uses the default compression level (Z_DEFAULT_COMPRESSION = 6). Tests
4261 have shown that for a large majority of images, compression values in
4262 the range 3-6 compress nearly as well as higher levels, and do so much
4263 faster. For online applications it may be desirable to have maximum speed
4264 (Z_BEST_SPEED = 1). With versions of zlib after v0.99, you can also
4265 specify no compression (Z_NO_COMPRESSION = 0), but this would create
4266 files larger than just storing the raw bitmap. You can specify the
4267 compression level by calling:
4270 png_set_compression_level(png_ptr, level);
4272 Another useful one is to reduce the memory level used by the library.
4273 The memory level defaults to 8, but it can be lowered if you are
4274 short on memory (running DOS, for example, where you only have 640K).
4275 Note that the memory level does have an effect on compression; among
4276 other things, lower levels will result in sections of incompressible
4277 data being emitted in smaller stored blocks, with a correspondingly
4278 larger relative overhead of up to 15% in the worst case.
4281 png_set_compression_mem_level(png_ptr, level);
4283 The other functions are for configuring zlib. They are not recommended
4284 for normal use and may result in writing an invalid PNG file. See
4285 zlib.h for more information on what these mean.
4288 png_set_compression_strategy(png_ptr,
4291 png_set_compression_window_bits(png_ptr,
4294 png_set_compression_method(png_ptr, method);
4296 This controls the size of the IDAT chunks (default 8192):
4298 png_set_compression_buffer_size(png_ptr, size);
4300 As of libpng version 1.5.4, additional APIs became
4301 available to set these separately for non-IDAT
4302 compressed chunks such as zTXt, iTXt, and iCCP:
4305 #if PNG_LIBPNG_VER >= 10504
4306 png_set_text_compression_level(png_ptr, level);
4308 png_set_text_compression_mem_level(png_ptr, level);
4310 png_set_text_compression_strategy(png_ptr,
4313 png_set_text_compression_window_bits(png_ptr,
4316 png_set_text_compression_method(png_ptr, method);
4319 Controlling row filtering
4321 If you want to control whether libpng uses filtering or not, which
4322 filters are used, and how it goes about picking row filters, you
4323 can call one of these functions. The selection and configuration
4324 of row filters can have a significant impact on the size and
4325 encoding speed and a somewhat lesser impact on the decoding speed
4326 of an image. Filtering is enabled by default for RGB and grayscale
4327 images (with and without alpha), but not for paletted images nor
4328 for any images with bit depths less than 8 bits/pixel.
4330 The 'method' parameter sets the main filtering method, which is
4331 currently only '0' in the PNG 1.2 specification. The 'filters'
4332 parameter sets which filter(s), if any, should be used for each
4333 scanline. Possible values are PNG_ALL_FILTERS and PNG_NO_FILTERS
4334 to turn filtering on and off, respectively.
4336 Individual filter types are PNG_FILTER_NONE, PNG_FILTER_SUB,
4337 PNG_FILTER_UP, PNG_FILTER_AVG, PNG_FILTER_PAETH, which can be bitwise
4338 ORed together with '|' to specify one or more filters to use.
4339 These filters are described in more detail in the PNG specification.
4340 If you intend to change the filter type during the course of writing
4341 the image, you should start with flags set for all of the filters
4342 you intend to use so that libpng can initialize its internal
4343 structures appropriately for all of the filter types. (Note that this
4344 means the first row must always be adaptively filtered, because libpng
4345 currently does not allocate the filter buffers until png_write_row()
4346 is called for the first time.)
4348 filters = PNG_FILTER_NONE | PNG_FILTER_SUB
4349 PNG_FILTER_UP | PNG_FILTER_AVG |
4350 PNG_FILTER_PAETH | PNG_ALL_FILTERS;
4352 png_set_filter(png_ptr, PNG_FILTER_TYPE_BASE,
4354 The second parameter can also be
4355 PNG_INTRAPIXEL_DIFFERENCING if you are
4356 writing a PNG to be embedded in a MNG
4357 datastream. This parameter must be the
4358 same as the value of filter_method used
4361 It is also possible to influence how libpng chooses from among the
4362 available filters. This is done in one or both of two ways - by
4363 telling it how important it is to keep the same filter for successive
4364 rows, and by telling it the relative computational costs of the filters.
4366 double weights[3] = {1.5, 1.3, 1.1},
4367 costs[PNG_FILTER_VALUE_LAST] =
4368 {1.0, 1.3, 1.3, 1.5, 1.7};
4370 png_set_filter_heuristics(png_ptr,
4371 PNG_FILTER_HEURISTIC_WEIGHTED, 3,
4374 The weights are multiplying factors that indicate to libpng that the
4375 row filter should be the same for successive rows unless another row filter
4376 is that many times better than the previous filter. In the above example,
4377 if the previous 3 filters were SUB, SUB, NONE, the SUB filter could have a
4378 "sum of absolute differences" 1.5 x 1.3 times higher than other filters
4379 and still be chosen, while the NONE filter could have a sum 1.1 times
4380 higher than other filters and still be chosen. Unspecified weights are
4381 taken to be 1.0, and the specified weights should probably be declining
4382 like those above in order to emphasize recent filters over older filters.
4384 The filter costs specify for each filter type a relative decoding cost
4385 to be considered when selecting row filters. This means that filters
4386 with higher costs are less likely to be chosen over filters with lower
4387 costs, unless their "sum of absolute differences" is that much smaller.
4388 The costs do not necessarily reflect the exact computational speeds of
4389 the various filters, since this would unduly influence the final image
4392 Note that the numbers above were invented purely for this example and
4393 are given only to help explain the function usage. Little testing has
4394 been done to find optimum values for either the costs or the weights.
4396 Requesting debug printout
4398 The macro definition PNG_DEBUG can be used to request debugging
4399 printout. Set it to an integer value in the range 0 to 3. Higher
4400 numbers result in increasing amounts of debugging information. The
4401 information is printed to the "stderr" file, unless another file
4402 name is specified in the PNG_DEBUG_FILE macro definition.
4404 When PNG_DEBUG > 0, the following functions (macros) become available:
4406 png_debug(level, message)
4407 png_debug1(level, message, p1)
4408 png_debug2(level, message, p1, p2)
4410 in which "level" is compared to PNG_DEBUG to decide whether to print
4411 the message, "message" is the formatted string to be printed,
4412 and p1 and p2 are parameters that are to be embedded in the string
4413 according to printf-style formatting directives. For example,
4415 png_debug1(2, "foo=%d", foo);
4420 fprintf(PNG_DEBUG_FILE, "foo=%d\n", foo);
4422 When PNG_DEBUG is defined but is zero, the macros aren't defined, but you
4423 can still use PNG_DEBUG to control your own debugging:
4429 When PNG_DEBUG = 1, the macros are defined, but only png_debug statements
4430 having level = 0 will be printed. There aren't any such statements in
4431 this version of libpng, but if you insert some they will be printed.
4435 The MNG specification (available at http://www.libpng.org/pub/mng) allows
4436 certain extensions to PNG for PNG images that are embedded in MNG datastreams.
4437 Libpng can support some of these extensions. To enable them, use the
4438 png_permit_mng_features() function:
4440 feature_set = png_permit_mng_features(png_ptr, mask)
4442 mask is a png_uint_32 containing the bitwise OR of the
4443 features you want to enable. These include
4444 PNG_FLAG_MNG_EMPTY_PLTE
4445 PNG_FLAG_MNG_FILTER_64
4446 PNG_ALL_MNG_FEATURES
4448 feature_set is a png_uint_32 that is the bitwise AND of
4449 your mask with the set of MNG features that is
4450 supported by the version of libpng that you are using.
4452 It is an error to use this function when reading or writing a standalone
4453 PNG file with the PNG 8-byte signature. The PNG datastream must be wrapped
4454 in a MNG datastream. As a minimum, it must have the MNG 8-byte signature
4455 and the MHDR and MEND chunks. Libpng does not provide support for these
4456 or any other MNG chunks; your application must provide its own support for
4457 them. You may wish to consider using libmng (available at
4458 http://www.libmng.com) instead.
4460 VIII. Changes to Libpng from version 0.88
4462 It should be noted that versions of libpng later than 0.96 are not
4463 distributed by the original libpng author, Guy Schalnat, nor by
4464 Andreas Dilger, who had taken over from Guy during 1996 and 1997, and
4465 distributed versions 0.89 through 0.96, but rather by another member
4466 of the original PNG Group, Glenn Randers-Pehrson. Guy and Andreas are
4467 still alive and well, but they have moved on to other things.
4469 The old libpng functions png_read_init(), png_write_init(),
4470 png_info_init(), png_read_destroy(), and png_write_destroy() have been
4471 moved to PNG_INTERNAL in version 0.95 to discourage their use. These
4472 functions will be removed from libpng version 1.4.0.
4474 The preferred method of creating and initializing the libpng structures is
4475 via the png_create_read_struct(), png_create_write_struct(), and
4476 png_create_info_struct() because they isolate the size of the structures
4477 from the application, allow version error checking, and also allow the
4478 use of custom error handling routines during the initialization, which
4479 the old functions do not. The functions png_read_destroy() and
4480 png_write_destroy() do not actually free the memory that libpng
4481 allocated for these structs, but just reset the data structures, so they
4482 can be used instead of png_destroy_read_struct() and
4483 png_destroy_write_struct() if you feel there is too much system overhead
4484 allocating and freeing the png_struct for each image read.
4486 Setting the error callbacks via png_set_message_fn() before
4487 png_read_init() as was suggested in libpng-0.88 is no longer supported
4488 because this caused applications that do not use custom error functions
4489 to fail if the png_ptr was not initialized to zero. It is still possible
4490 to set the error callbacks AFTER png_read_init(), or to change them with
4491 png_set_error_fn(), which is essentially the same function, but with a new
4492 name to force compilation errors with applications that try to use the old
4495 Support for the sCAL, iCCP, iTXt, and sPLT chunks was added at libpng-1.0.6;
4496 however, iTXt support was not enabled by default.
4498 Starting with version 1.0.7, you can find out which version of the library
4499 you are using at run-time:
4501 png_uint_32 libpng_vn = png_access_version_number();
4503 The number libpng_vn is constructed from the major version, minor
4504 version with leading zero, and release number with leading zero,
4505 (e.g., libpng_vn for version 1.0.7 is 10007).
4507 Note that this function does not take a png_ptr, so you can call it
4508 before you've created one.
4510 You can also check which version of png.h you used when compiling your
4513 png_uint_32 application_vn = PNG_LIBPNG_VER;
4515 IX. Changes to Libpng from version 1.0.x to 1.2.x
4517 Support for user memory management was enabled by default. To
4518 accomplish this, the functions png_create_read_struct_2(),
4519 png_create_write_struct_2(), png_set_mem_fn(), png_get_mem_ptr(),
4520 png_malloc_default(), and png_free_default() were added.
4522 Support for the iTXt chunk has been enabled by default as of
4525 Support for certain MNG features was enabled.
4527 Support for numbered error messages was added. However, we never got
4528 around to actually numbering the error messages. The function
4529 png_set_strip_error_numbers() was added (Note: the prototype for this
4530 function was inadvertently removed from png.h in PNG_NO_ASSEMBLER_CODE
4531 builds of libpng-1.2.15. It was restored in libpng-1.2.36).
4533 The png_malloc_warn() function was added at libpng-1.2.3. This issues
4534 a png_warning and returns NULL instead of aborting when it fails to
4535 acquire the requested memory allocation.
4537 Support for setting user limits on image width and height was enabled
4538 by default. The functions png_set_user_limits(), png_get_user_width_max(),
4539 and png_get_user_height_max() were added at libpng-1.2.6.
4541 The png_set_add_alpha() function was added at libpng-1.2.7.
4543 The function png_set_expand_gray_1_2_4_to_8() was added at libpng-1.2.9.
4544 Unlike png_set_gray_1_2_4_to_8(), the new function does not expand the
4545 tRNS chunk to alpha. The png_set_gray_1_2_4_to_8() function is
4548 A number of macro definitions in support of runtime selection of
4549 assembler code features (especially Intel MMX code support) were
4550 added at libpng-1.2.0:
4552 PNG_ASM_FLAG_MMX_SUPPORT_COMPILED
4553 PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU
4554 PNG_ASM_FLAG_MMX_READ_COMBINE_ROW
4555 PNG_ASM_FLAG_MMX_READ_INTERLACE
4556 PNG_ASM_FLAG_MMX_READ_FILTER_SUB
4557 PNG_ASM_FLAG_MMX_READ_FILTER_UP
4558 PNG_ASM_FLAG_MMX_READ_FILTER_AVG
4559 PNG_ASM_FLAG_MMX_READ_FILTER_PAETH
4560 PNG_ASM_FLAGS_INITIALIZED
4566 We added the following functions in support of runtime
4567 selection of assembler code features:
4569 png_get_mmx_flagmask()
4570 png_set_mmx_thresholds()
4572 png_get_mmx_bitdepth_threshold()
4573 png_get_mmx_rowbytes_threshold()
4576 We replaced all of these functions with simple stubs in libpng-1.2.20,
4577 when the Intel assembler code was removed due to a licensing issue.
4579 These macros are deprecated:
4581 PNG_READ_TRANSFORMS_NOT_SUPPORTED
4582 PNG_PROGRESSIVE_READ_NOT_SUPPORTED
4583 PNG_NO_SEQUENTIAL_READ_SUPPORTED
4584 PNG_WRITE_TRANSFORMS_NOT_SUPPORTED
4585 PNG_READ_ANCILLARY_CHUNKS_NOT_SUPPORTED
4586 PNG_WRITE_ANCILLARY_CHUNKS_NOT_SUPPORTED
4588 They have been replaced, respectively, by:
4590 PNG_NO_READ_TRANSFORMS
4591 PNG_NO_PROGRESSIVE_READ
4592 PNG_NO_SEQUENTIAL_READ
4593 PNG_NO_WRITE_TRANSFORMS
4594 PNG_NO_READ_ANCILLARY_CHUNKS
4595 PNG_NO_WRITE_ANCILLARY_CHUNKS
4597 PNG_MAX_UINT was replaced with PNG_UINT_31_MAX. It has been
4598 deprecated since libpng-1.0.16 and libpng-1.2.6.
4601 png_check_sig(sig, num)
4603 !png_sig_cmp(sig, 0, num)
4604 It has been deprecated since libpng-0.90.
4607 png_set_gray_1_2_4_to_8()
4608 which also expands tRNS to alpha was replaced with
4609 png_set_expand_gray_1_2_4_to_8()
4610 which does not. It has been deprecated since libpng-1.0.18 and 1.2.9.
4612 X. Changes to Libpng from version 1.0.x/1.2.x to 1.4.x
4614 Private libpng prototypes and macro definitions were moved from
4615 png.h and pngconf.h into a new pngpriv.h header file.
4617 Functions png_set_benign_errors(), png_benign_error(), and
4618 png_chunk_benign_error() were added.
4620 Support for setting the maximum amount of memory that the application
4621 will allocate for reading chunks was added, as a security measure.
4622 The functions png_set_chunk_cache_max() and png_get_chunk_cache_max()
4623 were added to the library.
4625 We implemented support for I/O states by adding png_ptr member io_state
4626 and functions png_get_io_chunk_name() and png_get_io_state() in pngget.c
4628 We added PNG_TRANSFORM_GRAY_TO_RGB to the available high-level
4631 Checking for and reporting of errors in the IHDR chunk is more thorough.
4633 Support for global arrays was removed, to improve thread safety.
4635 Some obsolete/deprecated macros and functions have been removed.
4637 Typecasted NULL definitions such as
4638 #define png_voidp_NULL (png_voidp)NULL
4639 were eliminated. If you used these in your application, just use
4642 The png_struct and info_struct members "trans" and "trans_values" were
4643 changed to "trans_alpha" and "trans_color", respectively.
4645 The obsolete, unused pnggccrd.c and pngvcrd.c files and related makefiles
4648 The PNG_1_0_X and PNG_1_2_X macros were eliminated.
4650 The PNG_LEGACY_SUPPORTED macro was eliminated.
4652 Many WIN32_WCE #ifdefs were removed.
4654 The functions png_read_init(info_ptr), png_write_init(info_ptr),
4655 png_info_init(info_ptr), png_read_destroy(), and png_write_destroy()
4656 have been removed. They have been deprecated since libpng-0.95.
4658 The png_permit_empty_plte() was removed. It has been deprecated
4659 since libpng-1.0.9. Use png_permit_mng_features() instead.
4661 We removed the obsolete stub functions png_get_mmx_flagmask(),
4662 png_set_mmx_thresholds(), png_get_asm_flags(),
4663 png_get_mmx_bitdepth_threshold(), png_get_mmx_rowbytes_threshold(),
4664 png_set_asm_flags(), and png_mmx_supported()
4666 We removed the obsolete png_check_sig(), png_memcpy_check(), and
4667 png_memset_check() functions. Instead use !png_sig_cmp(), memcpy(),
4668 and memset(), respectively.
4670 The function png_set_gray_1_2_4_to_8() was removed. It has been
4671 deprecated since libpng-1.0.18 and 1.2.9, when it was replaced with
4672 png_set_expand_gray_1_2_4_to_8() because the former function also
4673 expanded any tRNS chunk to an alpha channel.
4675 Macros for png_get_uint_16, png_get_uint_32, and png_get_int_32
4676 were added and are used by default instead of the corresponding
4677 functions. Unfortunately,
4678 from libpng-1.4.0 until 1.4.4, the png_get_uint_16 macro (but not the
4679 function) incorrectly returned a value of type png_uint_32.
4681 We changed the prototype for png_malloc() from
4682 png_malloc(png_structp png_ptr, png_uint_32 size)
4684 png_malloc(png_structp png_ptr, png_alloc_size_t size)
4686 This also applies to the prototype for the user replacement malloc_fn().
4688 The png_calloc() function was added and is used in place of
4689 of "png_malloc(); memset();" except in the case in png_read_png()
4690 where the array consists of pointers; in this case a "for" loop is used
4691 after the png_malloc() to set the pointers to NULL, to give robust.
4692 behavior in case the application runs out of memory part-way through
4695 We changed the prototypes of png_get_compression_buffer_size() and
4696 png_set_compression_buffer_size() to work with png_size_t instead of
4699 Support for numbered error messages was removed by default, since we
4700 never got around to actually numbering the error messages. The function
4701 png_set_strip_error_numbers() was removed from the library by default.
4703 The png_zalloc() and png_zfree() functions are no longer exported.
4704 The png_zalloc() function no longer zeroes out the memory that it
4705 allocates. Applications that called png_zalloc(png_ptr, number, size)
4706 can call png_calloc(png_ptr, number*size) instead, and can call
4707 png_free() instead of png_zfree().
4709 Support for dithering was disabled by default in libpng-1.4.0, because
4710 it has not been well tested and doesn't actually "dither".
4712 removed, however, and could be enabled by building libpng with
4713 PNG_READ_DITHER_SUPPORTED defined. In libpng-1.4.2, this support
4714 was re-enabled, but the function was renamed png_set_quantize() to
4715 reflect more accurately what it actually does. At the same time,
4716 the PNG_DITHER_[RED,GREEN_BLUE]_BITS macros were also renamed to
4717 PNG_QUANTIZE_[RED,GREEN,BLUE]_BITS, and PNG_READ_DITHER_SUPPORTED
4718 was renamed to PNG_READ_QUANTIZE_SUPPORTED.
4720 We removed the trailing '.' from the warning and error messages.
4722 XI. Changes to Libpng from version 1.4.x to 1.5.x
4724 From libpng-1.4.0 until 1.4.4, the png_get_uint_16 macro (but not the
4725 function) incorrectly returned a value of type png_uint_32.
4726 The incorrect macro was removed from libpng-1.4.5.
4728 Checking for invalid palette index on write was added at libpng
4729 1.5.10. If a pixel contains an invalid (out-of-range) index libpng issues
4730 a benign error. This is enabled by default because this condition is an
4731 error according to the PNG specification, Clause 11.3.2, but the error can
4732 be ignored in each png_ptr with
4734 png_set_check_for_invalid_index(png_ptr, allowed);
4737 0: disable benign error (accept the
4738 invalid data without warning).
4739 1: enable benign error (treat the
4740 invalid data as an error or a
4743 If the error is ignored, or if png_benign_error() treats it as a warning,
4744 any invalid pixels are decoded as opaque black by the decoder and written
4745 as-is by the encoder.
4747 Retrieving the maximum palette index found was added at libpng-1.5.15.
4748 This statement must appear after png_read_png() or png_read_image() while
4749 reading, and after png_write_png() or png_write_image() while writing.
4751 int max_palette = png_get_palette_max(png_ptr, info_ptr);
4753 This will return the maximum palette index found in the image, or "-1" if
4754 the palette was not checked, or "0" if no palette was found. Note that this
4755 does not account for any palette index used by ancillary chunks such as the
4756 bKGD chunk; you must check those separately to determine the maximum
4757 palette index actually used.
4759 There are no substantial API changes between the non-deprecated parts of
4760 the 1.4.5 API and the 1.5.0 API; however, the ability to directly access
4761 members of the main libpng control structures, png_struct and png_info,
4762 deprecated in earlier versions of libpng, has been completely removed from
4765 We no longer include zlib.h in png.h. The include statement has been moved
4766 to pngstruct.h, where it is not accessible by applications. Applications that
4767 need access to information in zlib.h will need to add the '#include "zlib.h"'
4768 directive. It does not matter whether this is placed prior to or after
4769 the '"#include png.h"' directive.
4771 The png_sprintf(), png_strcpy(), and png_strncpy() macros are no longer used
4774 We moved the png_strlen(), png_memcpy(), png_memset(), and png_memcmp()
4775 macros into a private header file (pngpriv.h) that is not accessible to
4778 In png_get_iCCP, the type of "profile" was changed from png_charpp
4779 to png_bytepp, and in png_set_iCCP, from png_charp to png_const_bytep.
4781 There are changes of form in png.h, including new and changed macros to
4782 declare parts of the API. Some API functions with arguments that are
4783 pointers to data not modified within the function have been corrected to
4784 declare these arguments with PNG_CONST.
4786 Much of the internal use of C macros to control the library build has also
4787 changed and some of this is visible in the exported header files, in
4788 particular the use of macros to control data and API elements visible
4789 during application compilation may require significant revision to
4790 application code. (It is extremely rare for an application to do this.)
4792 Any program that compiled against libpng 1.4 and did not use deprecated
4793 features or access internal library structures should compile and work
4794 against libpng 1.5, except for the change in the prototype for
4795 png_get_iCCP() and png_set_iCCP() API functions mentioned above.
4797 libpng 1.5.0 adds PNG_ PASS macros to help in the reading and writing of
4798 interlaced images. The macros return the number of rows and columns in
4799 each pass and information that can be used to de-interlace and (if
4800 absolutely necessary) interlace an image.
4802 libpng 1.5.0 adds an API png_longjmp(png_ptr, value). This API calls
4803 the application-provided png_longjmp_ptr on the internal, but application
4804 initialized, longjmp buffer. It is provided as a convenience to avoid
4805 the need to use the png_jmpbuf macro, which had the unnecessary side
4806 effect of resetting the internal png_longjmp_ptr value.
4808 libpng 1.5.0 includes a complete fixed point API. By default this is
4809 present along with the corresponding floating point API. In general the
4810 fixed point API is faster and smaller than the floating point one because
4811 the PNG file format used fixed point, not floating point. This applies
4812 even if the library uses floating point in internal calculations. A new
4813 macro, PNG_FLOATING_ARITHMETIC_SUPPORTED, reveals whether the library
4814 uses floating point arithmetic (the default) or fixed point arithmetic
4815 internally for performance critical calculations such as gamma correction.
4816 In some cases, the gamma calculations may produce slightly different
4817 results. This has changed the results in png_rgb_to_gray and in alpha
4818 composition (png_set_background for example). This applies even if the
4819 original image was already linear (gamma == 1.0) and, therefore, it is
4820 not necessary to linearize the image. This is because libpng has *not*
4821 been changed to optimize that case correctly, yet.
4823 Fixed point support for the sCAL chunk comes with an important caveat;
4824 the sCAL specification uses a decimal encoding of floating point values
4825 and the accuracy of PNG fixed point values is insufficient for
4826 representation of these values. Consequently a "string" API
4827 (png_get_sCAL_s and png_set_sCAL_s) is the only reliable way of reading
4828 arbitrary sCAL chunks in the absence of either the floating point API or
4829 internal floating point calculations. Starting with libpng-1.5.0, both
4830 of these functions are present when PNG_sCAL_SUPPORTED is defined. Prior
4831 to libpng-1.5.0, their presence also depended upon PNG_FIXED_POINT_SUPPORTED
4832 being defined and PNG_FLOATING_POINT_SUPPORTED not being defined.
4834 Applications no longer need to include the optional distribution header
4835 file pngusr.h or define the corresponding macros during application
4836 build in order to see the correct variant of the libpng API. From 1.5.0
4837 application code can check for the corresponding _SUPPORTED macro:
4839 #ifdef PNG_INCH_CONVERSIONS_SUPPORTED
4840 /* code that uses the inch conversion APIs. */
4843 This macro will only be defined if the inch conversion functions have been
4844 compiled into libpng. The full set of macros, and whether or not support
4845 has been compiled in, are available in the header file pnglibconf.h.
4846 This header file is specific to the libpng build. Notice that prior to
4847 1.5.0 the _SUPPORTED macros would always have the default definition unless
4848 reset by pngusr.h or by explicit settings on the compiler command line.
4849 These settings may produce compiler warnings or errors in 1.5.0 because
4850 of macro redefinition.
4852 Applications can now choose whether to use these macros or to call the
4853 corresponding function by defining PNG_USE_READ_MACROS or
4854 PNG_NO_USE_READ_MACROS before including png.h. Notice that this is
4855 only supported from 1.5.0; defining PNG_NO_USE_READ_MACROS prior to 1.5.0
4856 will lead to a link failure.
4858 Prior to libpng-1.5.4, the zlib compressor used the same set of parameters
4859 when compressing the IDAT data and textual data such as zTXt and iCCP.
4860 In libpng-1.5.4 we reinitialized the zlib stream for each type of data.
4861 We added five png_set_text_*() functions for setting the parameters to
4862 use with textual data.
4864 Prior to libpng-1.5.4, the PNG_READ_16_TO_8_ACCURATE_SCALE_SUPPORTED
4865 option was off by default, and slightly inaccurate scaling occurred.
4866 This option can no longer be turned off, and the choice of accurate
4867 or inaccurate 16-to-8 scaling is by using the new png_set_scale_16_to_8()
4868 API for accurate scaling or the old png_set_strip_16_to_8() API for simple
4869 chopping. In libpng-1.5.4, the PNG_READ_16_TO_8_ACCURATE_SCALE_SUPPORTED
4870 macro became PNG_READ_SCALE_16_TO_8_SUPPORTED, and the PNG_READ_16_TO_8
4871 macro became PNG_READ_STRIP_16_TO_8_SUPPORTED, to enable the two
4872 png_set_*_16_to_8() functions separately.
4874 Prior to libpng-1.5.4, the png_set_user_limits() function could only be
4875 used to reduce the width and height limits from the value of
4876 PNG_USER_WIDTH_MAX and PNG_USER_HEIGHT_MAX, although this document said
4877 that it could be used to override them. Now this function will reduce or
4878 increase the limits.
4880 Starting in libpng-1.5.10, the user limits can be set en masse with the
4881 configuration option PNG_SAFE_LIMITS_SUPPORTED. If this option is enabled,
4882 a set of "safe" limits is applied in pngpriv.h. These can be overridden by
4883 application calls to png_set_user_limits(), png_set_user_chunk_cache_max(),
4884 and/or png_set_user_malloc_max() that increase or decrease the limits. Also,
4885 in libpng-1.5.10 the default width and height limits were increased
4886 from 1,000,000 to 0x7ffffff (i.e., made unlimited). Therefore, the
4889 png_user_width_max 0x7fffffff 1,000,000
4890 png_user_height_max 0x7fffffff 1,000,000
4891 png_user_chunk_cache_max 0 (unlimited) 128
4892 png_user_chunk_malloc_max 0 (unlimited) 8,000,000
4894 The png_set_option() function (and the "options" member of the png struct) was
4895 added to libpng-1.5.15.
4897 The library now supports a complete fixed point implementation and can
4898 thus be used on systems that have no floating point support or very
4899 limited or slow support. Previously gamma correction, an essential part
4900 of complete PNG support, required reasonably fast floating point.
4902 As part of this the choice of internal implementation has been made
4903 independent of the choice of fixed versus floating point APIs and all the
4904 missing fixed point APIs have been implemented.
4906 The exact mechanism used to control attributes of API functions has
4907 changed, as described in the INSTALL file.
4909 A new test program, pngvalid, is provided in addition to pngtest.
4910 pngvalid validates the arithmetic accuracy of the gamma correction
4911 calculations and includes a number of validations of the file format.
4912 A subset of the full range of tests is run when "make check" is done
4913 (in the 'configure' build.) pngvalid also allows total allocated memory
4914 usage to be evaluated and performs additional memory overwrite validation.
4916 Many changes to individual feature macros have been made. The following
4917 are the changes most likely to be noticed by library builders who
4920 1) All feature macros now have consistent naming:
4922 #define PNG_NO_feature turns the feature off
4923 #define PNG_feature_SUPPORTED turns the feature on
4925 pnglibconf.h contains one line for each feature macro which is either:
4927 #define PNG_feature_SUPPORTED
4929 if the feature is supported or:
4931 /*#undef PNG_feature_SUPPORTED*/
4933 if it is not. Library code consistently checks for the 'SUPPORTED' macro.
4934 It does not, and libpng applications should not, check for the 'NO' macro
4935 which will not normally be defined even if the feature is not supported.
4936 The 'NO' macros are only used internally for setting or not setting the
4937 corresponding 'SUPPORTED' macros.
4939 Compatibility with the old names is provided as follows:
4941 PNG_INCH_CONVERSIONS turns on PNG_INCH_CONVERSIONS_SUPPORTED
4943 And the following definitions disable the corresponding feature:
4945 PNG_SETJMP_NOT_SUPPORTED disables SETJMP
4946 PNG_READ_TRANSFORMS_NOT_SUPPORTED disables READ_TRANSFORMS
4947 PNG_NO_READ_COMPOSITED_NODIV disables READ_COMPOSITE_NODIV
4948 PNG_WRITE_TRANSFORMS_NOT_SUPPORTED disables WRITE_TRANSFORMS
4949 PNG_READ_ANCILLARY_CHUNKS_NOT_SUPPORTED disables READ_ANCILLARY_CHUNKS
4950 PNG_WRITE_ANCILLARY_CHUNKS_NOT_SUPPORTED disables WRITE_ANCILLARY_CHUNKS
4952 Library builders should remove use of the above, inconsistent, names.
4954 2) Warning and error message formatting was previously conditional on
4955 the STDIO feature. The library has been changed to use the
4956 CONSOLE_IO feature instead. This means that if CONSOLE_IO is disabled
4957 the library no longer uses the printf(3) functions, even though the
4958 default read/write implementations use (FILE) style stdio.h functions.
4960 3) Three feature macros now control the fixed/floating point decisions:
4962 PNG_FLOATING_POINT_SUPPORTED enables the floating point APIs
4964 PNG_FIXED_POINT_SUPPORTED enables the fixed point APIs; however, in
4965 practice these are normally required internally anyway (because the PNG
4966 file format is fixed point), therefore in most cases PNG_NO_FIXED_POINT
4967 merely stops the function from being exported.
4969 PNG_FLOATING_ARITHMETIC_SUPPORTED chooses between the internal floating
4970 point implementation or the fixed point one. Typically the fixed point
4971 implementation is larger and slower than the floating point implementation
4972 on a system that supports floating point; however, it may be faster on a
4973 system which lacks floating point hardware and therefore uses a software
4976 4) Added PNG_{READ,WRITE}_INT_FUNCTIONS_SUPPORTED. This allows the
4977 functions to read and write ints to be disabled independently of
4978 PNG_USE_READ_MACROS, which allows libpng to be built with the functions
4979 even though the default is to use the macros - this allows applications
4980 to choose at app buildtime whether or not to use macros (previously
4981 impossible because the functions weren't in the default build.)
4983 XII. Changes to Libpng from version 1.5.x to 1.6.x
4985 A "simplified API" has been added (see documentation in png.h and a simple
4986 example in contrib/examples/pngtopng.c). The new publicly visible API
4987 includes the following:
4996 png_image_begin_read_from_file()
4997 png_image_begin_read_from_stdio()
4998 png_image_begin_read_from_memory()
4999 png_image_finish_read()
5002 png_image_write_to_file()
5003 png_image_write_to_stdio()
5005 Starting with libpng-1.6.0, you can configure libpng to prefix all exported
5006 symbols, using the PNG_PREFIX macro.
5008 We no longer include string.h in png.h. The include statement has been moved
5009 to pngpriv.h, where it is not accessible by applications. Applications that
5010 need access to information in string.h must add an '#include <string.h>'
5011 directive. It does not matter whether this is placed prior to or after
5012 the '#include "png.h"' directive.
5014 The following API are now DEPRECATED:
5016 png_convert_to_rfc1123() which has been replaced
5017 with png_convert_to_rfc1123_buffer()
5018 png_malloc_default()
5022 The following have been removed:
5023 png_get_io_chunk_name(), which has been replaced
5024 with png_get_io_chunk_type(). The new
5025 function returns a 32-bit integer instead of
5027 The png_sizeof(), png_strlen(), png_memcpy(), png_memcmp(), and
5028 png_memset() macros are no longer used in the libpng sources and
5029 have been removed. These had already been made invisible to applications
5030 (i.e., defined in the private pngpriv.h header file) since libpng-1.5.0.
5032 The signatures of many exported functions were changed, such that
5033 png_structp became png_structrp or png_const_structrp
5034 png_infop became png_inforp or png_const_inforp
5035 where "rp" indicates a "restricted pointer".
5037 Error detection in some chunks has improved; in particular the iCCP chunk
5038 reader now does pretty complete validation of the basic format. Some bad
5039 profiles that were previously accepted are now accepted with a warning or
5040 rejected, depending upon the png_set_benign_errors() setting, in particular the
5041 very old broken Microsoft/HP 3144-byte sRGB profile. Starting with
5042 libpng-1.6.11, recognizing and checking sRGB profiles can be avoided by
5045 #if defined(PNG_SKIP_sRGB_CHECK_PROFILE) && \
5046 defined(PNG_SET_OPTION_SUPPORTED)
5047 png_set_option(png_ptr, PNG_SKIP_sRGB_CHECK_PROFILE,
5051 It's not a good idea to do this if you are using the "simplified API",
5052 which needs to be able to recognize an sRGB profile conveyed via the iCCP
5055 The PNG spec requirement that only grayscale profiles may appear in images
5056 with color type 0 or 4 and that even if the image only contains gray pixels,
5057 only RGB profiles may appear in images with color type 2, 3, or 6, is now
5058 enforced. The sRGB chunk is allowed to appear in images with any color type
5059 and is interpreted by libpng to convey a one-tracer-curve gray profile or a
5060 three-tracer-curve RGB profile as appropriate.
5062 Prior to libpng-1.6.0 a warning would be issued if the iTXt chunk contained
5063 an empty language field or an empty translated keyword. Both of these
5064 are allowed by the PNG specification, so these warnings are no longer issued.
5066 The library now issues an error if the application attempts to set a
5067 transform after it calls png_read_update_info() or if it attempts to call
5068 both png_read_update_info() and png_start_read_image() or to call either
5069 of them more than once.
5071 The default condition for benign_errors is now to treat benign errors as
5072 warnings while reading and as errors while writing.
5074 The library now issues a warning if both background processing and RGB to
5075 gray are used when gamma correction happens. As with previous versions of
5076 the library the results are numerically very incorrect in this case.
5078 There are some minor arithmetic changes in some transforms such as
5079 png_set_background(), that might be detected by certain regression tests.
5081 Unknown chunk handling has been improved internally, without any API change.
5082 This adds more correct option control of the unknown handling, corrects
5083 a pre-existing bug where the per-chunk 'keep' setting is ignored, and makes
5084 it possible to skip IDAT chunks in the sequential reader.
5086 The machine-generated configure files are no longer included in branches
5087 libpng16 and later of the GIT repository. They continue to be included
5088 in the tarball releases, however.
5090 Libpng-1.6.0 through 1.6.2 used the CMF bytes at the beginning of the IDAT
5091 stream to set the size of the sliding window for reading instead of using the
5092 default 32-kbyte sliding window size. It was discovered that there are
5093 hundreds of PNG files in the wild that have incorrect CMF bytes that caused
5094 zlib to issue the "invalid distance too far back" error and reject the file.
5095 Libpng-1.6.3 and later calculate their own safe CMF from the image dimensions,
5096 provide a way to revert to the libpng-1.5.x behavior (ignoring the CMF bytes
5097 and using a 32-kbyte sliding window), by using
5099 png_set_option(png_ptr, PNG_MAXIMUM_INFLATE_WINDOW,
5102 and provide a tool (contrib/tools/pngfix) for rewriting a PNG file while
5103 optimizing the CMF bytes in its IDAT chunk correctly.
5105 Libpng-1.6.0 and libpng-1.6.1 wrote uncompressed iTXt chunks with the wrong
5106 length, which resulted in PNG files that cannot be read beyond the bad iTXt
5107 chunk. This error was fixed in libpng-1.6.3, and a tool (called
5108 contrib/tools/png-fix-itxt) has been added to the libpng distribution.
5110 XIII. Detecting libpng
5112 The png_get_io_ptr() function has been present since libpng-0.88, has never
5113 changed, and is unaffected by conditional compilation macros. It is the
5114 best choice for use in configure scripts for detecting the presence of any
5115 libpng version since 0.88. In an autoconf "configure.in" you could use
5117 AC_CHECK_LIB(png, png_get_io_ptr, ...
5119 XV. Source code repository
5121 Since about February 2009, version 1.2.34, libpng has been under "git" source
5122 control. The git repository was built from old libpng-x.y.z.tar.gz files
5123 going back to version 0.70. You can access the git repository (read only)
5126 git://git.code.sf.net/p/libpng/code
5128 or you can browse it with a web browser by selecting the "code" button at
5130 https://sourceforge.net/projects/libpng
5132 Patches can be sent to glennrp at users.sourceforge.net or to
5133 png-mng-implement at lists.sourceforge.net or you can upload them to
5134 the libpng bug tracker at
5136 http://libpng.sourceforge.net
5138 We also accept patches built from the tar or zip distributions, and
5139 simple verbal discriptions of bug fixes, reported either to the
5140 SourceForge bug tracker, to the png-mng-implement at lists.sf.net
5141 mailing list, or directly to glennrp.
5145 Our coding style is similar to the "Allman" style
5146 (See http://en.wikipedia.org/wiki/Indent_style#Allman_style), with curly
5147 braces on separate lines:
5154 else if (another condition)
5159 The braces can be omitted from simple one-line actions:
5164 We use 3-space indentation, except for continued statements which
5165 are usually indented the same as the first line of the statement
5166 plus four more spaces.
5168 For macro definitions we use 2-space indentation, always leaving the "#"
5169 in the first column.
5171 #ifndef PNG_NO_FEATURE
5172 # ifndef PNG_FEATURE_SUPPORTED
5173 # define PNG_FEATURE_SUPPORTED
5177 Comments appear with the leading "/*" at the same indentation as
5178 the statement that follows the comment:
5180 /* Single-line comment */
5183 /* This is a multiple-line
5188 Very short comments can be placed after the end of the statement
5189 to which they pertain:
5191 statement; /* comment */
5193 We don't use C++ style ("//") comments. We have, however,
5194 used them in the past in some now-abandoned MMX assembler
5197 Functions and their curly braces are not indented, and
5198 exported functions are marked with PNGAPI:
5200 /* This is a public function that is visible to
5201 * application programmers. It does thus-and-so.
5204 png_exported_function(png_ptr, png_info, foo)
5209 The return type and decorations are placed on a separate line
5210 ahead of the function name, as illustrated above.
5212 The prototypes for all exported functions appear in png.h,
5213 above the comment that says
5215 /* Maintainer: Put new public prototypes here ... */
5217 We mark all non-exported functions with "/* PRIVATE */"":
5220 png_non_exported_function(png_ptr, png_info, foo)
5225 The prototypes for non-exported functions (except for those in
5228 above the comment that says
5230 /* Maintainer: Put new private prototypes here ^ */
5232 We put a space after the "sizeof" operator and we omit the
5233 optional parentheses around its argument when the argument
5234 is an expression, not a type name, and we always enclose the
5235 sizeof operator, with its argument, in parentheses:
5237 (sizeof (png_uint_32))
5240 Prior to libpng-1.6.0 we used a "png_sizeof()" macro, formatted as
5241 though it were a function.
5243 To avoid polluting the global namespace, the names of all exported
5244 functions and variables begin with "png_", and all publicly visible C
5245 preprocessor macros begin with "PNG". We request that applications that
5246 use libpng *not* begin any of their own symbols with either of these strings.
5248 We put a space after each comma and after each semicolon
5249 in "for" statements, and we put spaces before and after each
5250 C binary operator and after "for" or "while", and before
5251 "?". We don't put a space between a typecast and the expression
5252 being cast, nor do we put one between a function name and the
5253 left parenthesis that follows it:
5255 for (i = 2; i > 0; --i)
5256 y[i] = a(x) + (int)b;
5258 We prefer #ifdef and #ifndef to #if defined() and #if !defined()
5259 when there is only one macro being tested. We always use parentheses
5262 We prefer to express integers that are used as bit masks in hex format,
5263 with an even number of lower-case hex digits (e.g., 0x00, 0xff, 0x0100).
5265 We prefer to use underscores in variable names rather than camelCase, except
5266 for a few type names that we inherit from zlib.h.
5268 We prefer "if (something != 0)" and "if (something == 0)"
5269 over "if (something)" and if "(!something)", respectively.
5271 We do not use the TAB character for indentation in the C sources.
5273 Lines do not exceed 80 characters.
5275 Other rules can be inferred by inspecting the libpng source.
5277 XVI. Y2K Compliance in libpng
5281 Since the PNG Development group is an ad-hoc body, we can't make
5282 an official declaration.
5284 This is your unofficial assurance that libpng from version 0.71 and
5285 upward through 1.6.13 are Y2K compliant. It is my belief that earlier
5286 versions were also Y2K compliant.
5288 Libpng only has two year fields. One is a 2-byte unsigned integer
5289 that will hold years up to 65535. The other, which is deprecated,
5290 holds the date in text format, and will hold years up to 9999.
5293 "png_uint_16 year" in png_time_struct.
5296 "char time_buffer[29]" in png_struct. This is no longer used
5297 in libpng-1.6.x and will be removed from libpng-1.7.0.
5299 There are seven time-related functions:
5301 png_convert_to_rfc_1123() in png.c
5302 (formerly png_convert_to_rfc_1152() in error)
5303 png_convert_from_struct_tm() in pngwrite.c, called
5305 png_convert_from_time_t() in pngwrite.c
5306 png_get_tIME() in pngget.c
5307 png_handle_tIME() in pngrutil.c, called in pngread.c
5308 png_set_tIME() in pngset.c
5309 png_write_tIME() in pngwutil.c, called in pngwrite.c
5311 All appear to handle dates properly in a Y2K environment. The
5312 png_convert_from_time_t() function calls gmtime() to convert from system
5313 clock time, which returns (year - 1900), which we properly convert to
5314 the full 4-digit year. There is a possibility that applications using
5315 libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
5316 function, or that they are incorrectly passing only a 2-digit year
5317 instead of "year - 1900" into the png_convert_from_struct_tm() function,
5318 but this is not under our control. The libpng documentation has always
5319 stated that it works with 4-digit years, and the APIs have been
5322 The tIME chunk itself is also Y2K compliant. It uses a 2-byte unsigned
5323 integer to hold the year, and can hold years as large as 65535.
5325 zlib, upon which libpng depends, is also Y2K compliant. It contains
5326 no date-related code.
5329 Glenn Randers-Pehrson
5331 PNG Development Group