1 USING THE IJG JPEG LIBRARY
3 This file was part of the Independent JPEG Group's software:
4 Copyright (C) 1994-2013, Thomas G. Lane, Guido Vollbeding.
5 Lossless JPEG Modifications:
6 Copyright (C) 1999, Ken Murchison.
7 libjpeg-turbo Modifications:
8 Copyright (C) 2010, 2014-2018, 2020, 2022-2023, D. R. Commander.
9 Copyright (C) 2015, Google, Inc.
10 For conditions of distribution and use, see the accompanying README.ijg file.
13 This file describes how to use the IJG JPEG library within an application
14 program. Read it if you want to write a program that uses the library.
16 The file example.c provides heavily commented code for calling the JPEG
17 library. Also see jpeglib.h (the include file to be used by application
18 programs) for full details about data structures and function parameter lists.
19 The library source code, of course, is the ultimate reference.
21 Note that there have been *major* changes from the application interface
22 presented by IJG version 4 and earlier versions. The old design had several
23 inherent limitations, and it had accumulated a lot of cruft as we added
24 features while trying to minimize application-interface changes. We have
25 sacrificed backward compatibility in the version 5 rewrite, but we think the
26 improvements justify this.
33 Functions provided by the library
34 12-bit and 16-bit Data Precision
35 Outline of typical usage
40 Partial image decompression
41 Mechanics of usage: include files, linking, etc
43 Compression parameter selection
44 Decompression parameter selection
47 Compressed data handling (source and destination managers)
49 Progressive JPEG support
51 Abbreviated datastreams and multiple images
54 Raw (downsampled) image data
55 Really raw data: DCT coefficients
59 Library compile-time options
60 Portability considerations
62 You should read at least the overview and basic usage sections before trying
63 to program with the library. The sections on advanced features can be read
64 if and when you need them.
70 Functions provided by the library
71 ---------------------------------
73 The IJG JPEG library provides C code to read and write JPEG-compressed image
74 files. The surrounding application program receives or supplies image data a
75 scanline at a time, using a straightforward uncompressed image format. All
76 details of color conversion and other preprocessing/postprocessing can be
77 handled by the library.
79 The library includes a substantial amount of code that is not covered by the
80 JPEG standard but is necessary for typical applications of JPEG. These
81 functions preprocess the image before JPEG compression or postprocess it after
82 decompression. They include colorspace conversion, downsampling/upsampling,
83 and color quantization. The application indirectly selects use of this code
84 by specifying the format in which it wishes to supply or receive image data.
85 For example, if colormapped output is requested, then the decompression
86 library automatically invokes color quantization.
88 A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
89 and even more so in decompression postprocessing. The decompression library
90 provides multiple implementations that cover most of the useful tradeoffs,
91 ranging from very-high-quality down to fast-preview operation. On the
92 compression side we have generally not provided low-quality choices, since
93 compression is normally less time-critical. It should be understood that the
94 low-quality modes may not meet the JPEG standard's accuracy requirements;
95 nonetheless, they are useful for viewers.
97 A word about functions *not* provided by the library. We handle a subset of
98 the ISO JPEG standard; most baseline, extended-sequential, and progressive
99 JPEG processes are supported. (Our subset includes all features now in common
100 use.) Unsupported ISO options include:
101 * Hierarchical storage
103 * Nonintegral subsampling ratios
104 We support 8-bit (lossy and lossless), 12-bit (lossy and lossless), and 16-bit
105 (lossless) data precision.
107 By itself, the library handles only interchange JPEG datastreams --- in
108 particular the widely used JFIF file format. The library can be used by
109 surrounding code to process interchange or abbreviated JPEG datastreams that
110 are embedded in more complex file formats. (For example, this library is
111 used by the free LIBTIFF library to support JPEG compression in TIFF.)
114 12-bit and 16-bit Data Precision
115 --------------------------------
117 The JPEG standard provides for baseline 8-bit and 12-bit DCT processes as well
118 as 8-bit, 12-bit, and 16-bit lossless (predictive) processes. This code
119 supports 12-bit-per-component lossy or lossless JPEG if you set
120 cinfo->data_precision to 12 and 16-bit-per-component lossless JPEG if you set
121 cinfo->data_precision to 16. Note that this causes the sample size to be
122 larger than a char, so it affects the surrounding application's image data.
123 The sample applications cjpeg and djpeg can support 12-bit mode only for PPM,
124 PGM, and GIF file formats and 16-bit mode only for PPM and PGM file formats.
126 Note that, when 12-bit data precision is enabled, the library always compresses
127 in Huffman optimization mode, in order to generate valid Huffman tables. This
128 is necessary because our default Huffman tables only cover 8-bit data. If you
129 need to output 12-bit files in one pass, you'll have to supply suitable default
130 Huffman tables. You may also want to supply your own DCT quantization tables;
131 the existing quality-scaling code has been developed for 8-bit use, and
132 probably doesn't generate especially good tables for 12-bit.
134 Functions that are specific to 12-bit data precision have a prefix of "jpeg12_"
135 instead of "jpeg_" and use the following data types and macros:
137 * J12SAMPLE instead of JSAMPLE
138 * J12SAMPROW instead of JSAMPROW
139 * J12SAMPARRAY instead of JSAMPARRAY
140 * J12SAMPIMAGE instead of JSAMPIMAGE
141 * MAXJ12SAMPLE instead of MAXJSAMPLE
142 * CENTERJ12SAMPLE instead of CENTERJSAMPLE
144 Functions that are specific to 16-bit data precision have a prefix of "jpeg16_"
145 instead of "jpeg_" and use the following data types and macros:
147 * J16SAMPLE instead of JSAMPLE
148 * J16SAMPROW instead of JSAMPROW
149 * J16SAMPARRAY instead of JSAMPARRAY
150 * J16SAMPIMAGE instead of JSAMPIMAGE
151 * MAXJ16SAMPLE instead of MAXJSAMPLE
152 * CENTERJ16SAMPLE instead of CENTERJSAMPLE
154 This allows 8-bit, 12-bit, and 16-bit data precision to be used in a single
155 application. (Refer to example.c). Arithmetic coding and SIMD acceleration
156 are not currently implemented for 12-bit data precision, nor are they
157 implemented for lossless mode with any data precision.
159 Refer to the descriptions of the data_precision compression and decompression
160 parameters below for further information.
162 This documentation uses "J*SAMPLE", "J*SAMPROW", "J*SAMPARRAY", and
163 "J*SAMPIMAGE" to generically refer to the 8-bit, 12-bit, or 16-bit data types.
166 Outline of typical usage
167 ------------------------
169 The rough outline of a JPEG compression operation is:
171 Allocate and initialize a JPEG compression object
172 Specify the destination for the compressed data (eg, a file)
173 Set parameters for compression, including image size & colorspace
174 jpeg_start_compress(...);
175 while (scan lines remain to be written)
176 jpeg_write_scanlines(...); /* Use jpeg12_write_scanlines() for
177 12-bit data precision and
178 jpeg16_write_scanlines() for
179 16-bit data precision. */
180 jpeg_finish_compress(...);
181 Release the JPEG compression object
183 A JPEG compression object holds parameters and working state for the JPEG
184 library. We make creation/destruction of the object separate from starting
185 or finishing compression of an image; the same object can be re-used for a
186 series of image compression operations. This makes it easy to re-use the
187 same parameter settings for a sequence of images. Re-use of a JPEG object
188 also has important implications for processing abbreviated JPEG datastreams,
191 The image data to be compressed is supplied to jpeg*_write_scanlines() from
192 in-memory buffers. If the application is doing file-to-file compression,
193 reading image data from the source file is the application's responsibility.
194 The library emits compressed data by calling a "data destination manager",
195 which typically will write the data into a file; but the application can
196 provide its own destination manager to do something else.
198 Similarly, the rough outline of a JPEG decompression operation is:
200 Allocate and initialize a JPEG decompression object
201 Specify the source of the compressed data (eg, a file)
202 Call jpeg_read_header() to obtain image info
203 Set parameters for decompression
204 jpeg_start_decompress(...);
205 while (scan lines remain to be read)
206 jpeg_read_scanlines(...); /* Use jpeg12_read_scanlines() for
207 12-bit data precision and
208 jpeg16_read_scanlines() for
209 16-bit data precision. */
210 jpeg_finish_decompress(...);
211 Release the JPEG decompression object
213 This is comparable to the compression outline except that reading the
214 datastream header is a separate step. This is helpful because information
215 about the image's size, colorspace, etc is available when the application
216 selects decompression parameters. For example, the application can choose an
217 output scaling ratio that will fit the image into the available screen size.
219 The decompression library obtains compressed data by calling a data source
220 manager, which typically will read the data from a file; but other behaviors
221 can be obtained with a custom source manager. Decompressed data is delivered
222 into in-memory buffers passed to jpeg*_read_scanlines().
224 It is possible to abort an incomplete compression or decompression operation
225 by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
226 simply release it by calling jpeg_destroy().
228 JPEG compression and decompression objects are two separate struct types.
229 However, they share some common fields, and certain routines such as
230 jpeg_destroy() can work on either type of object.
232 The JPEG library has no static variables: all state is in the compression
233 or decompression object. Therefore it is possible to process multiple
234 compression and decompression operations concurrently, using multiple JPEG
237 Both compression and decompression can be done in an incremental memory-to-
238 memory fashion, if suitable source/destination managers are used. See the
239 section on "I/O suspension" for more details.
248 Before diving into procedural details, it is helpful to understand the
249 image data format that the JPEG library expects or returns.
251 The standard input image format is a rectangular array of pixels, with each
252 pixel having the same number of "component" or "sample" values (color
253 channels). You must specify how many components there are and the colorspace
254 interpretation of the components. Most applications will use RGB data
255 (three components per pixel) or grayscale data (one component per pixel).
256 PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
257 A remarkable number of people manage to miss this, only to find that their
258 programs don't work with grayscale JPEG files.
260 There is no provision for colormapped input. JPEG files are always full-color
261 or full grayscale (or sometimes another colorspace such as CMYK). You can
262 feed in a colormapped image by expanding it to full-color format. However
263 JPEG often doesn't work very well with source data that has been colormapped,
264 because of dithering noise. This is discussed in more detail in the JPEG FAQ
265 and the other references mentioned in the README.ijg file.
267 Pixels are stored by scanlines, with each scanline running from left to
268 right. The component values for each pixel are adjacent in the row; for
269 example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color. Each scanline is an
270 array of data type JSAMPLE or J12SAMPLE --- which is typically "unsigned char"
271 or "short" (respectively), unless you've changed jmorecfg.h. (You can also
272 change the RGB pixel layout, say to B,G,R order, by modifying jmorecfg.h. But
273 see the restrictions listed in that file before doing so.)
275 A 2-D array of pixels is formed by making a list of pointers to the starts of
276 scanlines; so the scanlines need not be physically adjacent in memory. Even
277 if you process just one scanline at a time, you must make a one-element
278 pointer array to conform to this structure. Pointers to J*SAMPLE rows are of
279 type J*SAMPROW, and the pointer to the pointer array is of type J*SAMPARRAY.
281 The library accepts or supplies one or more complete scanlines per call.
282 It is not possible to process part of a row at a time. Scanlines are always
283 processed top-to-bottom. You can process an entire image in one call if you
284 have it all in memory, but usually it's simplest to process one scanline at
287 For best results, source data values should have the precision specified by
288 cinfo->data_precision (normally 8 bits). For instance, if you choose to
289 compress data that's only 6 bits/channel, you should left-justify each value in
290 a byte before passing it to the compressor. If you need to compress data
291 that has more than 8 bits/channel, set cinfo->data_precision = 12 or 16.
294 The data format returned by the decompressor is the same in all details,
295 except that colormapped output is supported. (Again, a JPEG file is never
296 colormapped. But you can ask the decompressor to perform on-the-fly color
297 quantization to deliver colormapped output.) If you request colormapped
298 output then the returned data array contains a single J*SAMPLE per pixel;
299 its value is an index into a color map. The color map is represented as
300 a 2-D J*SAMPARRAY in which each row holds the values of one color component,
301 that is, colormap[i][j] is the value of the i'th color component for pixel
302 value (map index) j. Note that since the colormap indexes are stored in
303 J*SAMPLEs, the maximum number of colors is limited by the size of J*SAMPLE
304 (ie, at most 256 colors for 8-bit data precision, 4096 colors for 12-bit data
305 precision, and 65536 colors for 16-bit data precision).
311 Here we revisit the JPEG compression outline given in the overview.
313 1. Allocate and initialize a JPEG compression object.
315 A JPEG compression object is a "struct jpeg_compress_struct". (It also has
316 a bunch of subsidiary structures which are allocated via malloc(), but the
317 application doesn't control those directly.) This struct can be just a local
318 variable in the calling routine, if a single routine is going to execute the
319 whole JPEG compression sequence. Otherwise it can be static or allocated
322 You will also need a structure representing a JPEG error handler. The part
323 of this that the library cares about is a "struct jpeg_error_mgr". If you
324 are providing your own error handler, you'll typically want to embed the
325 jpeg_error_mgr struct in a larger structure; this is discussed later under
326 "Error handling". For now we'll assume you are just using the default error
327 handler. The default error handler will print JPEG error/warning messages
328 on stderr, and it will call exit() if a fatal error occurs.
330 You must initialize the error handler structure, store a pointer to it into
331 the JPEG object's "err" field, and then call jpeg_create_compress() to
332 initialize the rest of the JPEG object.
334 Typical code for this step, if you are using the default error handler, is
336 struct jpeg_compress_struct cinfo;
337 struct jpeg_error_mgr jerr;
339 cinfo.err = jpeg_std_error(&jerr);
340 jpeg_create_compress(&cinfo);
342 jpeg_create_compress allocates a small amount of memory, so it could fail
343 if you are out of memory. In that case it will exit via the error handler;
344 that's why the error handler must be initialized first.
347 2. Specify the destination for the compressed data (eg, a file).
349 As previously mentioned, the JPEG library delivers compressed data to a
350 "data destination" module. The library includes one data destination
351 module which knows how to write to a stdio stream. You can use your own
352 destination module if you want to do something else, as discussed later.
354 If you use the standard destination module, you must open the target stdio
355 stream beforehand. Typical code for this step looks like:
359 if ((outfile = fopen(filename, "wb")) == NULL) {
360 fprintf(stderr, "can't open %s\n", filename);
363 jpeg_stdio_dest(&cinfo, outfile);
365 where the last line invokes the standard destination module.
367 WARNING: it is critical that the binary compressed data be delivered to the
368 output file unchanged. On non-Unix systems the stdio library may perform
369 newline translation or otherwise corrupt binary data. To suppress this
370 behavior, you may need to use a "b" option to fopen (as shown above), or use
371 setmode() or another routine to put the stdio stream in binary mode. See
372 cjpeg.c and djpeg.c for code that has been found to work on many systems.
374 You can select the data destination after setting other parameters (step 3),
375 if that's more convenient. You may not change the destination between
376 calling jpeg_start_compress() and jpeg_finish_compress().
379 3. Set parameters for compression, including image size & colorspace.
381 You must supply information about the source image by setting the following
382 fields in the JPEG object (cinfo structure):
384 image_width Width of image, in pixels
385 image_height Height of image, in pixels
386 input_components Number of color channels (samples per pixel)
387 in_color_space Color space of source image
389 The image dimensions are, hopefully, obvious. JPEG supports image dimensions
390 of 1 to 64K pixels in either direction. The input color space is typically
391 RGB or grayscale, and input_components is 3 or 1 accordingly. (See "Special
392 color spaces", later, for more info.) The in_color_space field must be
393 assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
396 JPEG has a large number of compression parameters that determine how the
397 image is encoded. Most applications don't need or want to know about all
398 these parameters. You can set all the parameters to reasonable defaults by
399 calling jpeg_set_defaults(); then, if there are particular values you want
400 to change, you can do so after that. The "Compression parameter selection"
401 section tells about all the parameters.
403 You must set in_color_space correctly before calling jpeg_set_defaults(),
404 because the defaults depend on the source image colorspace. However the
405 other three source image parameters need not be valid until you call
406 jpeg_start_compress(). There's no harm in calling jpeg_set_defaults() more
407 than once, if that happens to be convenient.
409 Typical code for a 24-bit RGB source image is
411 cinfo.image_width = Width; /* image width and height, in pixels */
412 cinfo.image_height = Height;
413 cinfo.input_components = 3; /* # of color components per pixel */
414 cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
416 jpeg_set_defaults(&cinfo);
417 /* Make optional parameter settings here */
420 4. jpeg_start_compress(...);
422 After you have established the data destination and set all the necessary
423 source image info and other parameters, call jpeg_start_compress() to begin
424 a compression cycle. This will initialize internal state, allocate working
425 storage, and emit the first few bytes of the JPEG datastream header.
429 jpeg_start_compress(&cinfo, TRUE);
431 The "TRUE" parameter ensures that a complete JPEG interchange datastream
432 will be written. This is appropriate in most cases. If you think you might
433 want to use an abbreviated datastream, read the section on abbreviated
436 Once you have called jpeg_start_compress(), you may not alter any JPEG
437 parameters or other fields of the JPEG object until you have completed
438 the compression cycle.
441 5. while (scan lines remain to be written)
442 jpeg_write_scanlines(...); /* Use jpeg12_write_scanlines() for 12-bit
444 jpeg16_write_scanlines() for 16-bit data
447 Now write all the required image data by calling jpeg*_write_scanlines()
448 one or more times. You can pass one or more scanlines in each call, up
449 to the total image height. In most applications it is convenient to pass
450 just one or a few scanlines at a time. The expected format for the passed
451 data is discussed under "Data formats", above.
453 Image data should be written in top-to-bottom scanline order.
454 Rec. ITU-T T.81 | ISO/IEC 10918-1 says, "Applications determine which edges of
455 a source image are defined as top, bottom, left, and right." However, if you
456 want your files to be compatible with everyone else's, then top-to-bottom order
457 must be used. If the source data must be read in bottom-to-top order, then you
458 can use the JPEG library's virtual array mechanism to invert the data
459 efficiently. Examples of this can be found in the sample application cjpeg.
461 The library maintains a count of the number of scanlines written so far
462 in the next_scanline field of the JPEG object. Usually you can just use
463 this variable as the loop counter, so that the loop test looks like
464 "while (cinfo.next_scanline < cinfo.image_height)".
466 Code for this step depends heavily on the way that you store the source data.
467 example.c shows the following code for the case of a full-size 2-D source
468 array containing 3-byte RGB pixels:
470 JSAMPROW row_pointer[1]; /* pointer to a single row
471 Use J12SAMPROW for 12-bit data
472 precision and J16SAMPROW for 16-bit
475 while (cinfo.next_scanline < cinfo.image_height) {
476 row_pointer[0] = image_buffer[cinfo.next_scanline];
477 jpeg_write_scanlines(&cinfo, row_pointer, 1);
478 /* Use jpeg12_write_scanlines() for
479 12-bit data precision and
480 jpeg16_write_scanlines() for 16-bit
484 jpeg*_write_scanlines() returns the number of scanlines actually written.
485 This will normally be equal to the number passed in, so you can usually
486 ignore the return value. It is different in just two cases:
487 * If you try to write more scanlines than the declared image height,
488 the additional scanlines are ignored.
489 * If you use a suspending data destination manager, output buffer overrun
490 will cause the compressor to return before accepting all the passed lines.
491 This feature is discussed under "I/O suspension", below. The normal
492 stdio destination manager will NOT cause this to happen.
493 In any case, the return value is the same as the change in the value of
497 6. jpeg_finish_compress(...);
499 After all the image data has been written, call jpeg_finish_compress() to
500 complete the compression cycle. This step is ESSENTIAL to ensure that the
501 last bufferload of data is written to the data destination.
502 jpeg_finish_compress() also releases working memory associated with the JPEG
507 jpeg_finish_compress(&cinfo);
509 If using the stdio destination manager, don't forget to close the output
510 stdio stream (if necessary) afterwards.
512 If you have requested a multi-pass operating mode, such as Huffman code
513 optimization, jpeg_finish_compress() will perform the additional passes using
514 data buffered by the first pass. In this case jpeg_finish_compress() may take
515 quite a while to complete. With the default compression parameters, this will
518 It is an error to call jpeg_finish_compress() before writing the necessary
519 total number of scanlines. If you wish to abort compression, call
520 jpeg_abort() as discussed below.
522 After completing a compression cycle, you may dispose of the JPEG object
523 as discussed next, or you may use it to compress another image. In that case
524 return to step 2, 3, or 4 as appropriate. If you do not change the
525 destination manager, the new datastream will be written to the same target.
526 If you do not change any JPEG parameters, the new datastream will be written
527 with the same parameters as before. Note that you can change the input image
528 dimensions freely between cycles, but if you change the input colorspace, you
529 should call jpeg_set_defaults() to adjust for the new colorspace; and then
530 you'll need to repeat all of step 3.
533 7. Release the JPEG compression object.
535 When you are done with a JPEG compression object, destroy it by calling
536 jpeg_destroy_compress(). This will free all subsidiary memory (regardless of
537 the previous state of the object). Or you can call jpeg_destroy(), which
538 works for either compression or decompression objects --- this may be more
539 convenient if you are sharing code between compression and decompression
540 cases. (Actually, these routines are equivalent except for the declared type
541 of the passed pointer. To avoid gripes from ANSI C compilers, jpeg_destroy()
542 should be passed a j_common_ptr.)
544 If you allocated the jpeg_compress_struct structure from malloc(), freeing
545 it is your responsibility --- jpeg_destroy() won't. Ditto for the error
550 jpeg_destroy_compress(&cinfo);
555 If you decide to abort a compression cycle before finishing, you can clean up
556 in either of two ways:
558 * If you don't need the JPEG object any more, just call
559 jpeg_destroy_compress() or jpeg_destroy() to release memory. This is
560 legitimate at any point after calling jpeg_create_compress() --- in fact,
561 it's safe even if jpeg_create_compress() fails.
563 * If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
564 jpeg_abort() which works on both compression and decompression objects.
565 This will return the object to an idle state, releasing any working memory.
566 jpeg_abort() is allowed at any time after successful object creation.
568 Note that cleaning up the data destination, if required, is your
569 responsibility; neither of these routines will call term_destination().
570 (See "Compressed data handling", below, for more about that.)
572 jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
573 object that has reported an error by calling error_exit (see "Error handling"
574 for more info). The internal state of such an object is likely to be out of
575 whack. Either of these two routines will return the object to a known state.
578 Decompression details
579 ---------------------
581 Here we revisit the JPEG decompression outline given in the overview.
583 1. Allocate and initialize a JPEG decompression object.
585 This is just like initialization for compression, as discussed above,
586 except that the object is a "struct jpeg_decompress_struct" and you
587 call jpeg_create_decompress(). Error handling is exactly the same.
591 struct jpeg_decompress_struct cinfo;
592 struct jpeg_error_mgr jerr;
594 cinfo.err = jpeg_std_error(&jerr);
595 jpeg_create_decompress(&cinfo);
597 (Both here and in the IJG code, we usually use variable name "cinfo" for
598 both compression and decompression objects.)
601 2. Specify the source of the compressed data (eg, a file).
603 As previously mentioned, the JPEG library reads compressed data from a "data
604 source" module. The library includes one data source module which knows how
605 to read from a stdio stream. You can use your own source module if you want
606 to do something else, as discussed later.
608 If you use the standard source module, you must open the source stdio stream
609 beforehand. Typical code for this step looks like:
613 if ((infile = fopen(filename, "rb")) == NULL) {
614 fprintf(stderr, "can't open %s\n", filename);
617 jpeg_stdio_src(&cinfo, infile);
619 where the last line invokes the standard source module.
621 WARNING: it is critical that the binary compressed data be read unchanged.
622 On non-Unix systems the stdio library may perform newline translation or
623 otherwise corrupt binary data. To suppress this behavior, you may need to use
624 a "b" option to fopen (as shown above), or use setmode() or another routine to
625 put the stdio stream in binary mode. See cjpeg.c and djpeg.c for code that
626 has been found to work on many systems.
628 You may not change the data source between calling jpeg_read_header() and
629 jpeg_finish_decompress(). If you wish to read a series of JPEG images from
630 a single source file, you should repeat the jpeg_read_header() to
631 jpeg_finish_decompress() sequence without reinitializing either the JPEG
632 object or the data source module; this prevents buffered input data from
636 3. Call jpeg_read_header() to obtain image info.
638 Typical code for this step is just
640 jpeg_read_header(&cinfo, TRUE);
642 This will read the source datastream header markers, up to the beginning
643 of the compressed data proper. On return, the image dimensions and other
644 info have been stored in the JPEG object. The application may wish to
645 consult this information before selecting decompression parameters.
647 More complex code is necessary if
648 * A suspending data source is used --- in that case jpeg_read_header()
649 may return before it has read all the header data. See "I/O suspension",
650 below. The normal stdio source manager will NOT cause this to happen.
651 * Abbreviated JPEG files are to be processed --- see the section on
652 abbreviated datastreams. Standard applications that deal only in
653 interchange JPEG files need not be concerned with this case either.
655 It is permissible to stop at this point if you just wanted to find out the
656 image dimensions and other header info for a JPEG file. In that case,
657 call jpeg_destroy() when you are done with the JPEG object, or call
658 jpeg_abort() to return it to an idle state before selecting a new data
659 source and reading another header.
662 4. Set parameters for decompression.
664 jpeg_read_header() sets appropriate default decompression parameters based on
665 the properties of the image (in particular, its colorspace). However, you
666 may well want to alter these defaults before beginning the decompression.
667 For example, the default is to produce full color output from a color file.
668 If you want colormapped output you must ask for it. Other options allow the
669 returned image to be scaled and allow various speed/quality tradeoffs to be
670 selected. "Decompression parameter selection", below, gives details.
672 If the defaults are appropriate, nothing need be done at this step.
674 Note that all default values are set by each call to jpeg_read_header().
675 If you reuse a decompression object, you cannot expect your parameter
676 settings to be preserved across cycles, as you can for compression.
677 You must set desired parameter values each time.
680 5. jpeg_start_decompress(...);
682 Once the parameter values are satisfactory, call jpeg_start_decompress() to
683 begin decompression. This will initialize internal state, allocate working
684 memory, and prepare for returning data.
688 jpeg_start_decompress(&cinfo);
690 If you have requested a multi-pass operating mode, such as 2-pass color
691 quantization, jpeg_start_decompress() will do everything needed before data
692 output can begin. In this case jpeg_start_decompress() may take quite a while
693 to complete. With a single-scan (non progressive) JPEG file and default
694 decompression parameters, this will not happen; jpeg_start_decompress() will
697 After this call, the final output image dimensions, including any requested
698 scaling, are available in the JPEG object; so is the selected colormap, if
699 colormapped output has been requested. Useful fields include
701 output_width image width and height, as scaled
703 out_color_components # of color components in out_color_space
704 output_components # of color components returned per pixel
705 colormap the selected colormap, if any
706 actual_number_of_colors number of entries in colormap
708 output_components is 1 (a colormap index) when quantizing colors; otherwise it
709 equals out_color_components. It is the number of J*SAMPLE values that will be
710 emitted per pixel in the output arrays.
712 Typically you will need to allocate data buffers to hold the incoming image.
713 You will need output_width * output_components J*SAMPLEs per scanline in your
714 output buffer, and a total of output_height scanlines will be returned.
716 Note: if you are using the JPEG library's internal memory manager to allocate
717 data buffers (as djpeg does), then the manager's protocol requires that you
718 request large buffers *before* calling jpeg_start_decompress(). This is a
719 little tricky since the output_XXX fields are not normally valid then. You
720 can make them valid by calling jpeg_calc_output_dimensions() after setting the
721 relevant parameters (scaling, output color space, and quantization flag).
724 6. while (scan lines remain to be read)
725 jpeg_read_scanlines(...); /* Use jpeg12_read_scanlines() for 12-bit
727 jpeg16_read_scanlines() for 16-bit data
730 Now you can read the decompressed image data by calling jpeg*_read_scanlines()
731 one or more times. At each call, you pass in the maximum number of scanlines
732 to be read (ie, the height of your working buffer); jpeg*_read_scanlines()
733 will return up to that many lines. The return value is the number of lines
734 actually read. The format of the returned data is discussed under "Data
735 formats", above. Don't forget that grayscale and color JPEGs will return
736 different data formats!
738 Image data is returned in top-to-bottom scanline order. If you must write
739 out the image in bottom-to-top order, you can use the JPEG library's virtual
740 array mechanism to invert the data efficiently. Examples of this can be
741 found in the sample application djpeg.
743 The library maintains a count of the number of scanlines returned so far
744 in the output_scanline field of the JPEG object. Usually you can just use
745 this variable as the loop counter, so that the loop test looks like
746 "while (cinfo.output_scanline < cinfo.output_height)". (Note that the test
747 should NOT be against image_height, unless you never use scaling. The
748 image_height field is the height of the original unscaled image.)
749 The return value always equals the change in the value of output_scanline.
751 If you don't use a suspending data source, it is safe to assume that
752 jpeg*_read_scanlines() reads at least one scanline per call, until the
753 bottom of the image has been reached.
755 If you use a buffer larger than one scanline, it is NOT safe to assume that
756 jpeg*_read_scanlines() fills it. (The current implementation returns only a
757 few scanlines per call, no matter how large a buffer you pass.) So you must
758 always provide a loop that calls jpeg*_read_scanlines() repeatedly until the
759 whole image has been read.
762 7. jpeg_finish_decompress(...);
764 After all the image data has been read, call jpeg_finish_decompress() to
765 complete the decompression cycle. This causes working memory associated
766 with the JPEG object to be released.
770 jpeg_finish_decompress(&cinfo);
772 If using the stdio source manager, don't forget to close the source stdio
775 It is an error to call jpeg_finish_decompress() before reading the correct
776 total number of scanlines. If you wish to abort decompression, call
777 jpeg_abort() as discussed below.
779 After completing a decompression cycle, you may dispose of the JPEG object as
780 discussed next, or you may use it to decompress another image. In that case
781 return to step 2 or 3 as appropriate. If you do not change the source
782 manager, the next image will be read from the same source.
785 8. Release the JPEG decompression object.
787 When you are done with a JPEG decompression object, destroy it by calling
788 jpeg_destroy_decompress() or jpeg_destroy(). The previous discussion of
789 destroying compression objects applies here too.
793 jpeg_destroy_decompress(&cinfo);
798 You can abort a decompression cycle by calling jpeg_destroy_decompress() or
799 jpeg_destroy() if you don't need the JPEG object any more, or
800 jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
801 The previous discussion of aborting compression cycles applies here too.
804 Partial image decompression
805 ---------------------------
807 Partial image decompression is convenient for performance-critical applications
808 that wish to view only a portion of a large JPEG image without decompressing
809 the whole thing. It it also useful in memory-constrained environments (such as
810 on mobile devices.) This library provides the following functions to support
811 partial image decompression:
813 1. Skipping rows when decompressing
815 jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines);
816 /* Use jpeg12_skip_scanlines() for 12-bit data precision. */
818 This function provides application programmers with the ability to skip over
819 multiple rows in the JPEG image.
821 Suspending data sources are not supported by this function. Calling
822 jpeg*_skip_scanlines() with a suspending data source will result in undefined
823 behavior. Two-pass color quantization is also not supported by this function.
824 Calling jpeg*_skip_scanlines() with two-pass color quantization enabled will
827 jpeg*_skip_scanlines() will not allow skipping past the bottom of the image.
828 If the value of num_lines is large enough to skip past the bottom of the image,
829 then the function will skip to the end of the image instead.
831 If the value of num_lines is valid, then jpeg*_skip_scanlines() will always
832 skip all of the input rows requested. There is no need to inspect the return
833 value of the function in that case.
835 Best results will be achieved by calling jpeg*_skip_scanlines() for large
836 chunks of rows. The function should be viewed as a way to quickly jump to a
837 particular vertical offset in the JPEG image in order to decode a subset of the
838 image. Used in this manner, it will provide significant performance
841 Calling jpeg*_skip_scanlines() for small values of num_lines has several
843 1) JPEG decompression occurs in blocks, so if jpeg*_skip_scanlines() is
844 called from the middle of a decompression block, then it is likely that
845 much of the decompression work has already been done for the first
846 couple of rows that need to be skipped.
847 2) When this function returns, it must leave the decompressor in a state
848 such that it is ready to read the next line. This may involve
849 decompressing a block that must be partially skipped.
850 These issues are especially tricky for cases in which upsampling requires
851 context rows. In the worst case, jpeg*_skip_scanlines() will perform similarly
852 to jpeg*_read_scanlines() (since it will actually call jpeg*_read_scanlines().)
854 2. Decompressing partial scanlines
856 jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
858 /* Use jpeg12_crop_scanline() for 12-bit data precision. */
860 This function provides application programmers with the ability to decompress
861 only a portion of each row in the JPEG image. It must be called after
862 jpeg_start_decompress() and before any calls to jpeg*_read_scanlines() or
863 jpeg*_skip_scanlines().
865 If xoffset and width do not form a valid subset of the image row, then this
866 function will generate an error. Note that if the output image is scaled, then
867 xoffset and width are relative to the scaled image dimensions.
869 xoffset and width are passed by reference because xoffset must fall on an iMCU
870 boundary. If it doesn't, then it will be moved left to the nearest iMCU
871 boundary, and width will be increased accordingly. If the calling program does
872 not like the adjusted values of xoffset and width, then it can call
873 jpeg*_crop_scanline() again with new values (for instance, if it wants to move
874 xoffset to the nearest iMCU boundary to the right instead of to the left.)
876 After calling this function, cinfo->output_width will be set to the adjusted
877 width. This value should be used when allocating an output buffer to pass to
878 jpeg*_read_scanlines().
880 The output image from a partial-width decompression will be identical to the
881 corresponding image region from a full decode, with one exception: The "fancy"
882 (smooth) h2v2 (4:2:0) and h2v1 (4:2:2) upsampling algorithms fill in the
883 missing chroma components by averaging the chroma components from neighboring
884 pixels, except on the right and left edges of the image (where there are no
885 neighboring pixels.) When performing a partial-width decompression, these
886 "fancy" upsampling algorithms may treat the left and right edges of the partial
887 image region as if they are the left and right edges of the image, meaning that
888 the upsampling algorithm may be simplified. The result is that the pixels on
889 the left or right edge of the partial image may not be exactly identical to the
890 corresponding pixels in the original image.
893 Mechanics of usage: include files, linking, etc
894 -----------------------------------------------
896 Applications using the JPEG library should include the header file jpeglib.h
897 to obtain declarations of data types and routines. Before including
898 jpeglib.h, include system headers that define at least the typedefs FILE and
899 size_t. On ANSI-conforming systems, including <stdio.h> is sufficient; on
900 older Unix systems, you may need <sys/types.h> to define size_t.
902 If the application needs to refer to individual JPEG library error codes, also
903 include jerror.h to define those symbols.
905 jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h. If you are
906 installing the JPEG header files in a system directory, you will want to
907 install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
909 The most convenient way to include the JPEG code into your executable program
910 is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
911 machines) and reference it at your link step. If you use only half of the
912 library (only compression or only decompression), only that much code will be
913 included from the library, unless your linker is hopelessly brain-damaged.
914 The supplied build system builds libjpeg.a automatically.
916 It may be worth pointing out that the core JPEG library does not actually
917 require the stdio library: only the default source/destination managers and
918 error handler need it. You can use the library in a stdio-less environment
919 if you replace those modules and use jmemnobs.c (or another memory manager of
920 your own devising). More info about the minimum system library requirements
921 may be found in jinclude.h.
927 Compression parameter selection
928 -------------------------------
930 This section describes all the optional parameters you can set for JPEG
931 compression, as well as the "helper" routines provided to assist in this
932 task. Proper setting of some parameters requires detailed understanding
933 of the JPEG standard; if you don't know what a parameter is for, it's best
934 not to mess with it! See REFERENCES in the README.ijg file for pointers to
935 more info about JPEG.
937 It's a good idea to call jpeg_set_defaults() first, even if you plan to set
938 all the parameters; that way your code is more likely to work with future JPEG
939 libraries that have additional parameters. For the same reason, we recommend
940 you use a helper routine where one is provided, in preference to twiddling
941 cinfo fields directly.
943 The helper routines are:
945 jpeg_set_defaults (j_compress_ptr cinfo)
946 This routine sets all JPEG parameters to reasonable defaults, using
947 only the input image's color space (field in_color_space, which must
948 already be set in cinfo). Many applications will only need to use
949 this routine and perhaps jpeg_set_quality().
951 jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
952 Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
953 and sets other color-space-dependent parameters appropriately. See
954 "Special color spaces", below, before using this. A large number of
955 parameters, including all per-component parameters, are set by this
956 routine; if you want to twiddle individual parameters you should call
957 jpeg_set_colorspace() before rather than after.
959 jpeg_default_colorspace (j_compress_ptr cinfo)
960 Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
961 and calls jpeg_set_colorspace(). This is actually a subroutine of
962 jpeg_set_defaults(). It's broken out in case you want to change
963 just the colorspace-dependent JPEG parameters.
965 jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
966 Constructs JPEG quantization tables appropriate for the indicated
967 quality setting. The quality value is expressed on the 0..100 scale
968 recommended by IJG (cjpeg's "-quality" switch uses this routine).
969 Note that the exact mapping from quality values to tables may change
970 in future IJG releases as more is learned about DCT quantization.
971 If the force_baseline parameter is TRUE, then the quantization table
972 entries are constrained to the range 1..255 for full JPEG baseline
973 compatibility. In the current implementation, this only makes a
974 difference for quality settings below 25, and it effectively prevents
975 very small/low quality files from being generated. The IJG decoder
976 is capable of reading the non-baseline files generated at low quality
977 settings when force_baseline is FALSE, but other decoders may not be.
979 jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
980 boolean force_baseline)
981 Same as jpeg_set_quality() except that the generated tables are the
982 sample tables given in Annex K (Clause K.1) of
983 Rec. ITU-T T.81 (1992) | ISO/IEC 10918-1:1994, multiplied by the
984 specified scale factor (which is expressed as a percentage; thus
985 scale_factor = 100 reproduces the spec's tables). Note that larger
986 scale factors give lower quality. This entry point is useful for
987 conforming to the Adobe PostScript DCT conventions, but we do not
988 recommend linear scaling as a user-visible quality scale otherwise.
989 force_baseline again constrains the computed table entries to 1..255.
991 int jpeg_quality_scaling (int quality)
992 Converts a value on the IJG-recommended quality scale to a linear
993 scaling percentage. Note that this routine may change or go away
994 in future releases --- IJG may choose to adopt a scaling method that
995 can't be expressed as a simple scalar multiplier, in which case the
996 premise of this routine collapses. Caveat user.
998 jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
999 [libjpeg v7+ API/ABI emulation only]
1000 Set default quantization tables with linear q_scale_factor[] values
1003 jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
1004 const unsigned int *basic_table,
1005 int scale_factor, boolean force_baseline)
1006 Allows an arbitrary quantization table to be created. which_tbl
1007 indicates which table slot to fill. basic_table points to an array
1008 of 64 unsigned ints given in normal array order. These values are
1009 multiplied by scale_factor/100 and then clamped to the range 1..65535
1010 (or to 1..255 if force_baseline is TRUE).
1011 CAUTION: prior to library version 6a, jpeg_add_quant_table expected
1012 the basic table to be given in JPEG zigzag order. If you need to
1013 write code that works with either older or newer versions of this
1014 routine, you must check the library version number. Something like
1015 "#if JPEG_LIB_VERSION >= 61" is the right test.
1017 jpeg_simple_progression (j_compress_ptr cinfo)
1018 Generates a default scan script for writing a progressive-JPEG file.
1019 This is the recommended method of creating a progressive file,
1020 unless you want to make a custom scan sequence. You must ensure that
1021 the JPEG color space is set correctly before calling this routine.
1023 jpeg_enable_lossless (j_compress_ptr cinfo, int predictor_selection_value,
1024 int point_transform)
1025 Enables lossless mode with the specified predictor selection value
1026 (1 - 7) and optional point transform (0 - {precision}-1, where
1027 {precision} is the JPEG data precision). A point transform value of 0
1028 is necessary in order to create a fully lossless JPEG image. (A
1029 non-zero point transform value right-shifts the input samples by the
1030 specified number of bits, which is effectively a form of lossy color
1031 quantization.) Note that the following features will be unavailable
1032 when compressing or decompressing lossless JPEG images:
1033 * Partial image decompression
1034 * Quality/quantization table selection
1035 * DCT/IDCT algorithm selection
1037 * Downsampling/upsampling
1038 * Color space conversion (the JPEG image will use the same color
1039 space as the input image)
1040 * Color quantization
1042 * Raw (downsampled) data input/output
1043 * Transcoding of DCT coefficients
1044 Any parameters used to enable or configure those features will be
1048 Compression parameters (cinfo fields) include:
1051 If TRUE, use arithmetic coding.
1052 If FALSE, use Huffman coding.
1055 To create a 12-bit-per-component JPEG file, set data_precision to 12
1056 prior to calling jpeg_start_compress() or using the memory manager,
1057 then use jpeg12_write_scanlines() or jpeg12_write_raw_data() instead of
1058 jpeg_write_scanlines() or jpeg_write_raw_data(). To create a
1059 16-bit-per-component lossless JPEG file, set data_precision to 16 prior
1060 to calling jpeg_start_compress() or using the memory manager, then use
1061 jpeg16_write_scanlines() instead of jpeg_write_scanlines().
1063 J_DCT_METHOD dct_method
1064 Selects the algorithm used for the DCT step. Choices are:
1065 JDCT_ISLOW: accurate integer method
1066 JDCT_IFAST: less accurate integer method [legacy feature]
1067 JDCT_FLOAT: floating-point method [legacy feature]
1068 JDCT_DEFAULT: default method (normally JDCT_ISLOW)
1069 JDCT_FASTEST: fastest method (normally JDCT_IFAST)
1070 When the Independent JPEG Group's software was first released in 1991,
1071 the compression time for a 1-megapixel JPEG image on a mainstream PC
1072 was measured in minutes. Thus, JDCT_IFAST provided noticeable
1073 performance benefits. On modern CPUs running libjpeg-turbo, however,
1074 the compression time for a 1-megapixel JPEG image is measured in
1075 milliseconds, and thus the performance benefits of JDCT_IFAST are much
1076 less noticeable. On modern x86/x86-64 CPUs that support AVX2
1077 instructions, JDCT_IFAST and JDCT_ISLOW have similar performance. On
1078 other types of CPUs, JDCT_IFAST is generally about 5-15% faster than
1081 For quality levels of 90 and below, there should be little or no
1082 perceptible quality difference between the two algorithms. For quality
1083 levels above 90, however, the difference between JDCT_IFAST and
1084 JDCT_ISLOW becomes more pronounced. With quality=97, for instance,
1085 JDCT_IFAST incurs generally about a 1-3 dB loss in PSNR relative to
1086 JDCT_ISLOW, but this can be larger for some images. Do not use
1087 JDCT_IFAST with quality levels above 97. The algorithm often
1088 degenerates at quality=98 and above and can actually produce a more
1089 lossy image than if lower quality levels had been used. Also, in
1090 libjpeg-turbo, JDCT_IFAST is not fully accelerated for quality levels
1091 above 97, so it will be slower than JDCT_ISLOW.
1093 JDCT_FLOAT does not produce significantly more accurate results than
1094 JDCT_ISLOW, and it is much slower. JDCT_FLOAT may also give different
1095 results on different machines due to varying roundoff behavior, whereas
1096 the integer methods should give the same results on all machines.
1098 J_COLOR_SPACE jpeg_color_space
1100 The JPEG color space and corresponding number of components; see
1101 "Special color spaces", below, for more info. We recommend using
1102 jpeg_set_color_space() if you want to change these.
1104 boolean optimize_coding
1105 TRUE causes the compressor to compute optimal Huffman coding tables
1106 for the image. This requires an extra pass over the data and
1107 therefore costs a good deal of space and time. The default is
1108 FALSE, which tells the compressor to use the supplied or default
1109 Huffman tables. In most cases optimal tables save only a few percent
1110 of file size compared to the default tables. Note that when this is
1111 TRUE, you need not supply Huffman tables at all, and any you do
1112 supply will be overwritten.
1114 unsigned int restart_interval
1116 To emit restart markers in the JPEG file, set one of these nonzero.
1117 Set restart_interval to specify the exact interval in MCU blocks
1118 (samples in lossless mode). Set restart_in_rows to specify the
1119 interval in MCU rows. (If restart_in_rows is not 0, then
1120 restart_interval is set after the image width in MCUs is computed.)
1121 Defaults are zero (no restarts). One restart marker per MCU row is
1122 often a good choice. NOTE: the overhead of restart markers is higher
1123 in grayscale JPEG files than in color files, and MUCH higher in
1124 progressive JPEGs. If you use restarts, you may want to use larger
1125 intervals in those cases.
1127 const jpeg_scan_info *scan_info
1129 By default, scan_info is NULL; this causes the compressor to write a
1130 single-scan sequential JPEG file. If not NULL, scan_info points to
1131 an array of scan definition records of length num_scans. The
1132 compressor will then write a JPEG file having one scan for each scan
1133 definition record. This is used to generate noninterleaved or
1134 progressive JPEG files. The library checks that the scan array
1135 defines a valid JPEG scan sequence. (jpeg_simple_progression creates
1136 a suitable scan definition array for progressive JPEG.) This is
1137 discussed further under "Progressive JPEG support".
1139 int smoothing_factor
1140 If non-zero, the input image is smoothed; the value should be 1 for
1141 minimal smoothing to 100 for maximum smoothing. Consult jcsample.c
1142 for details of the smoothing algorithm. The default is zero.
1144 boolean write_JFIF_header
1145 If TRUE, a JFIF APP0 marker is emitted. jpeg_set_defaults() and
1146 jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
1147 (ie, YCbCr or grayscale) is selected, otherwise FALSE.
1149 UINT8 JFIF_major_version
1150 UINT8 JFIF_minor_version
1151 The version number to be written into the JFIF marker.
1152 jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
1153 You should set it to 1.02 (major=1, minor=2) if you plan to write
1154 any JFIF 1.02 extension markers.
1159 The resolution information to be written into the JFIF marker;
1160 not used otherwise. density_unit may be 0 for unknown,
1161 1 for dots/inch, or 2 for dots/cm. The default values are 0,1,1
1162 indicating square pixels of unknown size.
1164 boolean write_Adobe_marker
1165 If TRUE, an Adobe APP14 marker is emitted. jpeg_set_defaults() and
1166 jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
1167 or YCCK is selected, otherwise FALSE. It is generally a bad idea
1168 to set both write_JFIF_header and write_Adobe_marker. In fact,
1169 you probably shouldn't change the default settings at all --- the
1170 default behavior ensures that the JPEG file's color space can be
1171 recognized by the decoder.
1173 JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS]
1174 Pointers to coefficient quantization tables, one per table slot,
1175 or NULL if no table is defined for a slot. Usually these should
1176 be set via one of the above helper routines; jpeg_add_quant_table()
1177 is general enough to define any quantization table. The other
1178 routines will set up table slot 0 for luminance quality and table
1179 slot 1 for chrominance.
1181 int q_scale_factor[NUM_QUANT_TBLS]
1182 [libjpeg v7+ API/ABI emulation only]
1183 Linear quantization scaling factors (0-100, default 100)
1184 for use with jpeg_default_qtables().
1185 See rdswitch.c and cjpeg.c for an example of usage.
1186 Note that the q_scale_factor[] values use "linear" scales, so JPEG
1187 quality levels chosen by the user must be converted to these scales
1188 using jpeg_quality_scaling(). Here is an example that corresponds to
1189 cjpeg -quality 90,70:
1191 jpeg_set_defaults(cinfo);
1193 /* Set luminance quality 90. */
1194 cinfo->q_scale_factor[0] = jpeg_quality_scaling(90);
1195 /* Set chrominance quality 70. */
1196 cinfo->q_scale_factor[1] = jpeg_quality_scaling(70);
1198 jpeg_default_qtables(cinfo, force_baseline);
1200 CAUTION: Setting separate quality levels for chrominance and luminance
1201 is mainly only useful if chrominance subsampling is disabled. 2x2
1202 chrominance subsampling (AKA "4:2:0") is the default, but you can
1203 explicitly disable subsampling as follows:
1205 cinfo->comp_info[0].v_samp_factor = 1;
1206 cinfo->comp_info[0].h_samp_factor = 1;
1208 JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
1209 JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
1210 Pointers to Huffman coding tables, one per table slot, or NULL if
1211 no table is defined for a slot. Slots 0 and 1 are filled with the
1212 JPEG sample tables by jpeg_set_defaults(). If you need to allocate
1213 more table structures, jpeg_alloc_huff_table() may be used.
1214 Note that optimal Huffman tables can be computed for an image
1215 by setting optimize_coding, as discussed above; there's seldom
1216 any need to mess with providing your own Huffman tables.
1219 [libjpeg v7+ API/ABI emulation only]
1220 The actual dimensions of the JPEG image that will be written to the file are
1221 given by the following fields. These are computed from the input image
1222 dimensions and the compression parameters by jpeg_start_compress(). You can
1223 also call jpeg_calc_jpeg_dimensions() to obtain the values that will result
1224 from the current parameter settings. This can be useful if you are trying
1225 to pick a scaling ratio that will get close to a desired target size.
1227 JDIMENSION jpeg_width Actual dimensions of output image.
1228 JDIMENSION jpeg_height
1231 Per-component parameters are stored in the struct cinfo.comp_info[i] for
1232 component number i. Note that components here refer to components of the
1233 JPEG color space, *not* the source image color space. A suitably large
1234 comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
1235 to use that routine, it's up to you to allocate the array.
1238 The one-byte identifier code to be recorded in the JPEG file for
1239 this component. For the standard color spaces, we recommend you
1240 leave the default values alone.
1244 Horizontal and vertical sampling factors for the component; must
1245 be 1..4 according to the JPEG standard. Note that larger sampling
1246 factors indicate a higher-resolution component; many people find
1247 this behavior quite unintuitive. The default values are 2,2 for
1248 luminance components and 1,1 for chrominance components, except
1249 for grayscale where 1,1 is used.
1252 Quantization table number for component. The default value is
1253 0 for luminance components and 1 for chrominance components.
1257 DC and AC entropy coding table numbers. The default values are
1258 0 for luminance components and 1 for chrominance components.
1261 Must equal the component's index in comp_info[]. (Beginning in
1262 release v6, the compressor library will fill this in automatically;
1266 Decompression parameter selection
1267 ---------------------------------
1269 Decompression parameter selection is somewhat simpler than compression
1270 parameter selection, since all of the JPEG internal parameters are
1271 recorded in the source file and need not be supplied by the application.
1272 (Unless you are working with abbreviated files, in which case see
1273 "Abbreviated datastreams", below.) Decompression parameters control
1274 the postprocessing done on the image to deliver it in a format suitable
1275 for the application's use. Many of the parameters control speed/quality
1276 tradeoffs, in which faster decompression may be obtained at the price of
1277 a poorer-quality image. The defaults select the highest quality (slowest)
1280 The following fields in the JPEG object are set by jpeg_read_header() and
1281 may be useful to the application in choosing decompression parameters:
1283 int data_precision Data precision (bits per component)
1284 If data_precision is 12, then use jpeg12_read_scanlines(),
1285 jpeg12_skip_scanlines(), jpeg12_crop_scanline(), and/or
1286 jpeg12_read_raw_data() instead of jpeg_read_scanlines(),
1287 jpeg_skip_scanlines(), jpeg_crop_scanline(), and/or
1288 jpeg_read_raw_data(). If data_precision is 16, then use
1289 jpeg16_read_scanlines() instead of jpeg_read_scanlines().
1291 JDIMENSION image_width Width and height of image
1292 JDIMENSION image_height
1293 int num_components Number of color components
1294 J_COLOR_SPACE jpeg_color_space Colorspace of image
1295 boolean saw_JFIF_marker TRUE if a JFIF APP0 marker was seen
1296 UINT8 JFIF_major_version Version information from JFIF marker
1297 UINT8 JFIF_minor_version
1298 UINT8 density_unit Resolution data from JFIF marker
1301 boolean saw_Adobe_marker TRUE if an Adobe APP14 marker was seen
1302 UINT8 Adobe_transform Color transform code from Adobe marker
1304 The JPEG color space, unfortunately, is something of a guess since the JPEG
1305 standard proper does not provide a way to record it. In practice most files
1306 adhere to the JFIF or Adobe conventions, and the decoder will recognize these
1307 correctly. See "Special color spaces", below, for more info.
1310 The decompression parameters that determine the basic properties of the
1313 J_COLOR_SPACE out_color_space
1314 Output color space. jpeg_read_header() sets an appropriate default
1315 based on jpeg_color_space; typically it will be RGB or grayscale.
1316 The application can change this field to request output in a different
1317 colorspace. For example, set it to JCS_GRAYSCALE to get grayscale
1318 output from a color file. (This is useful for previewing: grayscale
1319 output is faster than full color since the color components need not
1320 be processed.) Note that not all possible color space transforms are
1321 currently implemented; you may need to extend jdcolor.c if you want an
1324 unsigned int scale_num, scale_denom
1325 Scale the image by the fraction scale_num/scale_denom. Default is
1326 1/1, or no scaling. Currently, the only supported scaling ratios
1327 are M/8 with all M from 1 to 16, or any reduced fraction thereof (such
1328 as 1/2, 3/4, etc.) (The library design allows for arbitrary
1329 scaling ratios but this is not likely to be implemented any time soon.)
1330 Smaller scaling ratios permit significantly faster decoding since
1331 fewer pixels need be processed and a simpler IDCT method can be used.
1333 boolean quantize_colors
1334 If set TRUE, colormapped output will be delivered. Default is FALSE,
1335 meaning that full-color output will be delivered.
1337 The next three parameters are relevant only if quantize_colors is TRUE.
1339 int desired_number_of_colors
1340 Maximum number of colors to use in generating a library-supplied color
1341 map (the actual number of colors is returned in a different field).
1342 Default 256. Ignored when the application supplies its own color map.
1344 boolean two_pass_quantize
1345 If TRUE, an extra pass over the image is made to select a custom color
1346 map for the image. This usually looks a lot better than the one-size-
1347 fits-all colormap that is used otherwise. Default is TRUE. Ignored
1348 when the application supplies its own color map.
1350 J_DITHER_MODE dither_mode
1351 Selects color dithering method. Supported values are:
1352 JDITHER_NONE no dithering: fast, very low quality
1353 JDITHER_ORDERED ordered dither: moderate speed and quality
1354 JDITHER_FS Floyd-Steinberg dither: slow, high quality
1355 Default is JDITHER_FS. (At present, ordered dither is implemented
1356 only in the single-pass, standard-colormap case. If you ask for
1357 ordered dither when two_pass_quantize is TRUE or when you supply
1358 an external color map, you'll get F-S dithering.)
1360 When quantize_colors is TRUE, the target color map is described by the next
1361 two fields. colormap is set to NULL by jpeg_read_header(). The application
1362 can supply a color map by setting colormap non-NULL and setting
1363 actual_number_of_colors to the map size. Otherwise, jpeg_start_decompress()
1364 selects a suitable color map and sets these two fields itself.
1365 [Implementation restriction: at present, an externally supplied colormap is
1366 only accepted for 3-component output color spaces.]
1369 The color map, represented as a 2-D pixel array of out_color_components
1370 rows and actual_number_of_colors columns. Ignored if not quantizing.
1371 CAUTION: if the JPEG library creates its own colormap, the storage
1372 pointed to by this field is released by jpeg_finish_decompress().
1373 Copy the colormap somewhere else first, if you want to save it.
1374 CAUTION: if data_precision is 12 or 16, then this is actually a
1375 J12SAMPARRAY or a J16SAMPARRAY, so it must be type-cast in order to
1376 read/write 12-bit or 16-bit samples from/to the array.
1378 int actual_number_of_colors
1379 The number of colors in the color map.
1381 Additional decompression parameters that the application may set include:
1383 J_DCT_METHOD dct_method
1384 Selects the algorithm used for the DCT step. Choices are:
1385 JDCT_ISLOW: accurate integer method
1386 JDCT_IFAST: less accurate integer method [legacy feature]
1387 JDCT_FLOAT: floating-point method [legacy feature]
1388 JDCT_DEFAULT: default method (normally JDCT_ISLOW)
1389 JDCT_FASTEST: fastest method (normally JDCT_IFAST)
1390 When the Independent JPEG Group's software was first released in 1991,
1391 the decompression time for a 1-megapixel JPEG image on a mainstream PC
1392 was measured in minutes. Thus, JDCT_IFAST provided noticeable
1393 performance benefits. On modern CPUs running libjpeg-turbo, however,
1394 the decompression time for a 1-megapixel JPEG image is measured in
1395 milliseconds, and thus the performance benefits of JDCT_IFAST are much
1396 less noticeable. On modern x86/x86-64 CPUs that support AVX2
1397 instructions, JDCT_IFAST and JDCT_ISLOW have similar performance. On
1398 other types of CPUs, JDCT_IFAST is generally about 5-15% faster than
1401 If the JPEG image was compressed using a quality level of 85 or below,
1402 then there should be little or no perceptible quality difference
1403 between the two algorithms. When decompressing images that were
1404 compressed using quality levels above 85, however, the difference
1405 between JDCT_IFAST and JDCT_ISLOW becomes more pronounced. With images
1406 compressed using quality=97, for instance, JDCT_IFAST incurs generally
1407 about a 4-6 dB loss in PSNR relative to JDCT_ISLOW, but this can be
1408 larger for some images. If you can avoid it, do not use JDCT_IFAST
1409 when decompressing images that were compressed using quality levels
1410 above 97. The algorithm often degenerates for such images and can
1411 actually produce a more lossy output image than if the JPEG image had
1412 been compressed using lower quality levels.
1414 JDCT_FLOAT does not produce significantly more accurate results than
1415 JDCT_ISLOW, and it is much slower. JDCT_FLOAT may also give different
1416 results on different machines due to varying roundoff behavior, whereas
1417 the integer methods should give the same results on all machines.
1419 boolean do_fancy_upsampling
1420 If TRUE, do careful upsampling of chroma components. If FALSE,
1421 a faster but sloppier method is used. Default is TRUE. The visual
1422 impact of the sloppier method is often very small.
1424 boolean do_block_smoothing
1425 If TRUE, interblock smoothing is applied in early stages of decoding
1426 progressive JPEG files; if FALSE, not. Default is TRUE. Early
1427 progression stages look "fuzzy" with smoothing, "blocky" without.
1428 In any case, block smoothing ceases to be applied after the first few
1429 AC coefficients are known to full accuracy, so it is relevant only
1430 when using buffered-image mode for progressive images.
1432 boolean enable_1pass_quant
1433 boolean enable_external_quant
1434 boolean enable_2pass_quant
1435 These are significant only in buffered-image mode, which is
1436 described in its own section below.
1439 The output image dimensions are given by the following fields. These are
1440 computed from the source image dimensions and the decompression parameters
1441 by jpeg_start_decompress(). You can also call jpeg_calc_output_dimensions()
1442 to obtain the values that will result from the current parameter settings.
1443 This can be useful if you are trying to pick a scaling ratio that will get
1444 close to a desired target size. It's also important if you are using the
1445 JPEG library's memory manager to allocate output buffer space, because you
1446 are supposed to request such buffers *before* jpeg_start_decompress().
1448 JDIMENSION output_width Actual dimensions of output image.
1449 JDIMENSION output_height
1450 int out_color_components Number of color components in out_color_space.
1451 int output_components Number of color components returned.
1452 int rec_outbuf_height Recommended height of scanline buffer.
1454 When quantizing colors, output_components is 1, indicating a single color map
1455 index per pixel. Otherwise it equals out_color_components. The output arrays
1456 are required to be output_width * output_components J*SAMPLEs wide.
1458 rec_outbuf_height is the recommended minimum height (in scanlines) of the
1459 buffer passed to jpeg*_read_scanlines(). If the buffer is smaller, the
1460 library will still work, but time will be wasted due to unnecessary data
1461 copying. In high-quality modes, rec_outbuf_height is always 1, but some
1462 faster, lower-quality modes set it to larger values (typically 2 to 4).
1463 If you are going to ask for a high-speed processing mode, you may as well
1464 go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
1465 (An output buffer larger than rec_outbuf_height lines is OK, but won't
1466 provide any material speed improvement over that height.)
1469 Special color spaces
1470 --------------------
1472 The JPEG standard itself is "color blind" and doesn't specify any particular
1473 color space. It is customary to convert color data to a luminance/chrominance
1474 color space before compressing, since this permits greater compression. The
1475 existing de-facto JPEG file format standards specify YCbCr or grayscale data
1476 (JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe). For special
1477 applications such as multispectral images, other color spaces can be used,
1478 but it must be understood that such files will be unportable.
1480 The JPEG library can handle the most common colorspace conversions (namely
1481 RGB <=> YCbCr and CMYK <=> YCCK). It can also deal with data of an unknown
1482 color space, passing it through without conversion. If you deal extensively
1483 with an unusual color space, you can easily extend the library to understand
1484 additional color spaces and perform appropriate conversions.
1486 For compression, the source data's color space is specified by field
1487 in_color_space. This is transformed to the JPEG file's color space given
1488 by jpeg_color_space. jpeg_set_defaults() chooses a reasonable JPEG color
1489 space depending on in_color_space, but you can override this by calling
1490 jpeg_set_colorspace(). Of course you must select a supported transformation.
1491 jccolor.c currently supports the following transformations:
1496 plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
1497 YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
1499 The de-facto file format standards (JFIF and Adobe) specify APPn markers that
1500 indicate the color space of the JPEG file. It is important to ensure that
1501 these are written correctly, or omitted if the JPEG file's color space is not
1502 one of the ones supported by the de-facto standards. jpeg_set_colorspace()
1503 will set the compression parameters to include or omit the APPn markers
1504 properly, so long as it is told the truth about the JPEG color space.
1505 For example, if you are writing some random 3-component color space without
1506 conversion, don't try to fake out the library by setting in_color_space and
1507 jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN. You may want to write an
1508 APPn marker of your own devising to identify the colorspace --- see "Special
1511 When told that the color space is UNKNOWN, the library will default to using
1512 luminance-quality compression parameters for all color components. You may
1513 well want to change these parameters. See the source code for
1514 jpeg_set_colorspace(), in jcparam.c, for details.
1516 For decompression, the JPEG file's color space is given in jpeg_color_space,
1517 and this is transformed to the output color space out_color_space.
1518 jpeg_read_header's setting of jpeg_color_space can be relied on if the file
1519 conforms to JFIF or Adobe conventions, but otherwise it is no better than a
1520 guess. If you know the JPEG file's color space for certain, you can override
1521 jpeg_read_header's guess by setting jpeg_color_space. jpeg_read_header also
1522 selects a default output color space based on (its guess of) jpeg_color_space;
1523 set out_color_space to override this. Again, you must select a supported
1524 transformation. jdcolor.c currently supports
1530 as well as the null transforms. (Since GRAYSCALE=>RGB is provided, an
1531 application can force grayscale JPEGs to look like color JPEGs if it only
1532 wants to handle one case.)
1534 The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
1535 (it weights distances appropriately for RGB colors). You'll need to modify
1536 the code if you want to use it for non-RGB output color spaces. Note that
1537 jquant2.c is used to map to an application-supplied colormap as well as for
1538 the normal two-pass colormap selection process.
1540 CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
1541 files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
1542 This is arguably a bug in Photoshop, but if you need to work with Photoshop
1543 CMYK files, you will have to deal with it in your application. We cannot
1544 "fix" this in the library by inverting the data during the CMYK<=>YCCK
1545 transform, because that would break other applications, notably Ghostscript.
1546 Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
1547 data in the same inverted-YCCK representation used in bare JPEG files, but
1548 the surrounding PostScript code performs an inversion using the PS image
1549 operator. I am told that Photoshop 3.0 will write uninverted YCCK in
1550 EPS/JPEG files, and will omit the PS-level inversion. (But the data
1551 polarity used in bare JPEG files will not change in 3.0.) In either case,
1552 the JPEG library must not invert the data itself, or else Ghostscript would
1553 read these EPS files incorrectly.
1559 When the default error handler is used, any error detected inside the JPEG
1560 routines will cause a message to be printed on stderr, followed by exit().
1561 You can supply your own error handling routines to override this behavior
1562 and to control the treatment of nonfatal warnings and trace/debug messages.
1563 The file example.c illustrates the most common case, which is to have the
1564 application regain control after an error rather than exiting.
1566 The JPEG library never writes any message directly; it always goes through
1567 the error handling routines. Three classes of messages are recognized:
1568 * Fatal errors: the library cannot continue.
1569 * Warnings: the library can continue, but the data is corrupt, and a
1570 damaged output image is likely to result.
1571 * Trace/informational messages. These come with a trace level indicating
1572 the importance of the message; you can control the verbosity of the
1573 program by adjusting the maximum trace level that will be displayed.
1575 You may, if you wish, simply replace the entire JPEG error handling module
1576 (jerror.c) with your own code. However, you can avoid code duplication by
1577 only replacing some of the routines depending on the behavior you need.
1578 This is accomplished by calling jpeg_std_error() as usual, but then overriding
1579 some of the method pointers in the jpeg_error_mgr struct, as illustrated by
1582 All of the error handling routines will receive a pointer to the JPEG object
1583 (a j_common_ptr which points to either a jpeg_compress_struct or a
1584 jpeg_decompress_struct; if you need to tell which, test the is_decompressor
1585 field). This struct includes a pointer to the error manager struct in its
1586 "err" field. Frequently, custom error handler routines will need to access
1587 additional data which is not known to the JPEG library or the standard error
1588 handler. The most convenient way to do this is to embed either the JPEG
1589 object or the jpeg_error_mgr struct in a larger structure that contains
1590 additional fields; then casting the passed pointer provides access to the
1591 additional fields. Again, see example.c for one way to do it. (Beginning
1592 with IJG version 6b, there is also a void pointer "client_data" in each
1593 JPEG object, which the application can also use to find related data.
1594 The library does not touch client_data at all.)
1596 The individual methods that you might wish to override are:
1598 error_exit (j_common_ptr cinfo)
1599 Receives control for a fatal error. Information sufficient to
1600 generate the error message has been stored in cinfo->err; call
1601 output_message to display it. Control must NOT return to the caller;
1602 generally this routine will exit() or longjmp() somewhere.
1603 Typically you would override this routine to get rid of the exit()
1604 default behavior. Note that if you continue processing, you should
1605 clean up the JPEG object with jpeg_abort() or jpeg_destroy().
1607 output_message (j_common_ptr cinfo)
1608 Actual output of any JPEG message. Override this to send messages
1609 somewhere other than stderr. Note that this method does not know
1610 how to generate a message, only where to send it.
1612 format_message (j_common_ptr cinfo, char *buffer)
1613 Constructs a readable error message string based on the error info
1614 stored in cinfo->err. This method is called by output_message. Few
1615 applications should need to override this method. One possible
1616 reason for doing so is to implement dynamic switching of error message
1619 emit_message (j_common_ptr cinfo, int msg_level)
1620 Decide whether or not to emit a warning or trace message; if so,
1621 calls output_message. The main reason for overriding this method
1622 would be to abort on warnings. msg_level is -1 for warnings,
1623 0 and up for trace messages.
1625 Only error_exit() and emit_message() are called from the rest of the JPEG
1626 library; the other two are internal to the error handler.
1628 The actual message texts are stored in an array of strings which is pointed to
1629 by the field err->jpeg_message_table. The messages are numbered from 0 to
1630 err->last_jpeg_message, and it is these code numbers that are used in the
1631 JPEG library code. You could replace the message texts (for instance, with
1632 messages in French or German) by changing the message table pointer. See
1633 jerror.h for the default texts. CAUTION: this table will almost certainly
1634 change or grow from one library version to the next.
1636 It may be useful for an application to add its own message texts that are
1637 handled by the same mechanism. The error handler supports a second "add-on"
1638 message table for this purpose. To define an addon table, set the pointer
1639 err->addon_message_table and the message numbers err->first_addon_message and
1640 err->last_addon_message. If you number the addon messages beginning at 1000
1641 or so, you won't have to worry about conflicts with the library's built-in
1642 messages. See the sample applications cjpeg/djpeg for an example of using
1643 addon messages (the addon messages are defined in cderror.h).
1645 Actual invocation of the error handler is done via macros defined in jerror.h:
1646 ERREXITn(...) for fatal errors
1647 WARNMSn(...) for corrupt-data warnings
1648 TRACEMSn(...) for trace and informational messages.
1649 These macros store the message code and any additional parameters into the
1650 error handler struct, then invoke the error_exit() or emit_message() method.
1651 The variants of each macro are for varying numbers of additional parameters.
1652 The additional parameters are inserted into the generated message using
1653 standard printf() format codes.
1655 See jerror.h and jerror.c for further details.
1658 Compressed data handling (source and destination managers)
1659 ----------------------------------------------------------
1661 The JPEG compression library sends its compressed data to a "destination
1662 manager" module. The default destination manager just writes the data to a
1663 memory buffer or to a stdio stream, but you can provide your own manager to
1664 do something else. Similarly, the decompression library calls a "source
1665 manager" to obtain the compressed data; you can provide your own source
1666 manager if you want the data to come from somewhere other than a memory
1667 buffer or a stdio stream.
1669 In both cases, compressed data is processed a bufferload at a time: the
1670 destination or source manager provides a work buffer, and the library invokes
1671 the manager only when the buffer is filled or emptied. (You could define a
1672 one-character buffer to force the manager to be invoked for each byte, but
1673 that would be rather inefficient.) The buffer's size and location are
1674 controlled by the manager, not by the library. For example, the memory
1675 source manager just makes the buffer pointer and length point to the original
1676 data in memory. In this case the buffer-reload procedure will be invoked
1677 only if the decompressor ran off the end of the datastream, which would
1678 indicate an erroneous datastream.
1680 The work buffer is defined as an array of datatype JOCTET, which is generally
1681 "char" or "unsigned char". On a machine where char is not exactly 8 bits
1682 wide, you must define JOCTET as a wider data type and then modify the data
1683 source and destination modules to transcribe the work arrays into 8-bit units
1684 on external storage.
1686 A data destination manager struct contains a pointer and count defining the
1687 next byte to write in the work buffer and the remaining free space:
1689 JOCTET *next_output_byte; /* => next byte to write in buffer */
1690 size_t free_in_buffer; /* # of byte spaces remaining in buffer */
1692 The library increments the pointer and decrements the count until the buffer
1693 is filled. The manager's empty_output_buffer method must reset the pointer
1694 and count. The manager is expected to remember the buffer's starting address
1695 and total size in private fields not visible to the library.
1697 A data destination manager provides three methods:
1699 init_destination (j_compress_ptr cinfo)
1700 Initialize destination. This is called by jpeg_start_compress()
1701 before any data is actually written. It must initialize
1702 next_output_byte and free_in_buffer. free_in_buffer must be
1703 initialized to a positive value.
1705 empty_output_buffer (j_compress_ptr cinfo)
1706 This is called whenever the buffer has filled (free_in_buffer
1707 reaches zero). In typical applications, it should write out the
1708 *entire* buffer (use the saved start address and buffer length;
1709 ignore the current state of next_output_byte and free_in_buffer).
1710 Then reset the pointer & count to the start of the buffer, and
1711 return TRUE indicating that the buffer has been dumped.
1712 free_in_buffer must be set to a positive value when TRUE is
1713 returned. A FALSE return should only be used when I/O suspension is
1714 desired (this operating mode is discussed in the next section).
1716 term_destination (j_compress_ptr cinfo)
1717 Terminate destination --- called by jpeg_finish_compress() after all
1718 data has been written. In most applications, this must flush any
1719 data remaining in the buffer. Use either next_output_byte or
1720 free_in_buffer to determine how much data is in the buffer.
1722 term_destination() is NOT called by jpeg_abort() or jpeg_destroy(). If you
1723 want the destination manager to be cleaned up during an abort, you must do it
1726 You will also need code to create a jpeg_destination_mgr struct, fill in its
1727 method pointers, and insert a pointer to the struct into the "dest" field of
1728 the JPEG compression object. This can be done in-line in your setup code if
1729 you like, but it's probably cleaner to provide a separate routine similar to
1730 the jpeg_stdio_dest() or jpeg_mem_dest() routines of the supplied destination
1733 Decompression source managers follow a parallel design, but with some
1734 additional frammishes. The source manager struct contains a pointer and count
1735 defining the next byte to read from the work buffer and the number of bytes
1738 const JOCTET *next_input_byte; /* => next byte to read from buffer */
1739 size_t bytes_in_buffer; /* # of bytes remaining in buffer */
1741 The library increments the pointer and decrements the count until the buffer
1742 is emptied. The manager's fill_input_buffer method must reset the pointer and
1743 count. In most applications, the manager must remember the buffer's starting
1744 address and total size in private fields not visible to the library.
1746 A data source manager provides five methods:
1748 init_source (j_decompress_ptr cinfo)
1749 Initialize source. This is called by jpeg_read_header() before any
1750 data is actually read. Unlike init_destination(), it may leave
1751 bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
1752 will occur immediately).
1754 fill_input_buffer (j_decompress_ptr cinfo)
1755 This is called whenever bytes_in_buffer has reached zero and more
1756 data is wanted. In typical applications, it should read fresh data
1757 into the buffer (ignoring the current state of next_input_byte and
1758 bytes_in_buffer), reset the pointer & count to the start of the
1759 buffer, and return TRUE indicating that the buffer has been reloaded.
1760 It is not necessary to fill the buffer entirely, only to obtain at
1761 least one more byte. bytes_in_buffer MUST be set to a positive value
1762 if TRUE is returned. A FALSE return should only be used when I/O
1763 suspension is desired (this mode is discussed in the next section).
1765 skip_input_data (j_decompress_ptr cinfo, long num_bytes)
1766 Skip num_bytes worth of data. The buffer pointer and count should
1767 be advanced over num_bytes input bytes, refilling the buffer as
1768 needed. This is used to skip over a potentially large amount of
1769 uninteresting data (such as an APPn marker). In some applications
1770 it may be possible to optimize away the reading of the skipped data,
1771 but it's not clear that being smart is worth much trouble; large
1772 skips are uncommon. bytes_in_buffer may be zero on return.
1773 A zero or negative skip count should be treated as a no-op.
1775 resync_to_restart (j_decompress_ptr cinfo, int desired)
1776 This routine is called only when the decompressor has failed to find
1777 a restart (RSTn) marker where one is expected. Its mission is to
1778 find a suitable point for resuming decompression. For most
1779 applications, we recommend that you just use the default resync
1780 procedure, jpeg_resync_to_restart(). However, if you are able to back
1781 up in the input data stream, or if you have a-priori knowledge about
1782 the likely location of restart markers, you may be able to do better.
1783 Read the read_restart_marker() and jpeg_resync_to_restart() routines
1784 in jdmarker.c if you think you'd like to implement your own resync
1787 term_source (j_decompress_ptr cinfo)
1788 Terminate source --- called by jpeg_finish_decompress() after all
1789 data has been read. Often a no-op.
1791 For both fill_input_buffer() and skip_input_data(), there is no such thing
1792 as an EOF return. If the end of the file has been reached, the routine has
1793 a choice of exiting via ERREXIT() or inserting fake data into the buffer.
1794 In most cases, generating a warning message and inserting a fake EOI marker
1795 is the best course of action --- this will allow the decompressor to output
1796 however much of the image is there. In pathological cases, the decompressor
1797 may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
1798 jdatasrc.c illustrates the recommended error recovery behavior.
1800 term_source() is NOT called by jpeg_abort() or jpeg_destroy(). If you want
1801 the source manager to be cleaned up during an abort, you must do it yourself.
1803 You will also need code to create a jpeg_source_mgr struct, fill in its method
1804 pointers, and insert a pointer to the struct into the "src" field of the JPEG
1805 decompression object. This can be done in-line in your setup code if you
1806 like, but it's probably cleaner to provide a separate routine similar to the
1807 jpeg_stdio_src() or jpeg_mem_src() routines of the supplied source managers.
1809 For more information, consult the memory and stdio source and destination
1810 managers in jdatasrc.c and jdatadst.c.
1816 Some applications need to use the JPEG library as an incremental memory-to-
1817 memory filter: when the compressed data buffer is filled or emptied, they want
1818 control to return to the outer loop, rather than expecting that the buffer can
1819 be emptied or reloaded within the data source/destination manager subroutine.
1820 The library supports this need by providing an "I/O suspension" mode, which we
1821 describe in this section.
1823 The I/O suspension mode is not a panacea: nothing is guaranteed about the
1824 maximum amount of time spent in any one call to the library, so it will not
1825 eliminate response-time problems in single-threaded applications. If you
1826 need guaranteed response time, we suggest you "bite the bullet" and implement
1827 a real multi-tasking capability.
1829 To use I/O suspension, cooperation is needed between the calling application
1830 and the data source or destination manager; you will always need a custom
1831 source/destination manager. (Please read the previous section if you haven't
1832 already.) The basic idea is that the empty_output_buffer() or
1833 fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
1834 that it has done nothing. Upon seeing this, the JPEG library suspends
1835 operation and returns to its caller. The surrounding application is
1836 responsible for emptying or refilling the work buffer before calling the
1839 Compression suspension:
1841 For compression suspension, use an empty_output_buffer() routine that returns
1842 FALSE; typically it will not do anything else. This will cause the
1843 compressor to return to the caller of jpeg*_write_scanlines(), with the return
1844 value indicating that not all the supplied scanlines have been accepted.
1845 The application must make more room in the output buffer, adjust the output
1846 buffer pointer/count appropriately, and then call jpeg*_write_scanlines()
1847 again, pointing to the first unconsumed scanline.
1849 When forced to suspend, the compressor will backtrack to a convenient stopping
1850 point (usually the start of the current MCU); it will regenerate some output
1851 data when restarted. Therefore, although empty_output_buffer() is only
1852 called when the buffer is filled, you should NOT write out the entire buffer
1853 after a suspension. Write only the data up to the current position of
1854 next_output_byte/free_in_buffer. The data beyond that point will be
1855 regenerated after resumption.
1857 Because of the backtracking behavior, a good-size output buffer is essential
1858 for efficiency; you don't want the compressor to suspend often. (In fact, an
1859 overly small buffer could lead to infinite looping, if a single MCU required
1860 more data than would fit in the buffer.) We recommend a buffer of at least
1861 several Kbytes. You may want to insert explicit code to ensure that you don't
1862 call jpeg*_write_scanlines() unless there is a reasonable amount of space in
1863 the output buffer; in other words, flush the buffer before trying to compress
1866 The compressor does not allow suspension while it is trying to write JPEG
1867 markers at the beginning and end of the file. This means that:
1868 * At the beginning of a compression operation, there must be enough free
1869 space in the output buffer to hold the header markers (typically 600 or
1870 so bytes). The recommended buffer size is bigger than this anyway, so
1871 this is not a problem as long as you start with an empty buffer. However,
1872 this restriction might catch you if you insert large special markers, such
1873 as a JFIF thumbnail image, without flushing the buffer afterwards.
1874 * When you call jpeg_finish_compress(), there must be enough space in the
1875 output buffer to emit any buffered data and the final EOI marker. In the
1876 current implementation, half a dozen bytes should suffice for this, but
1877 for safety's sake we recommend ensuring that at least 100 bytes are free
1878 before calling jpeg_finish_compress().
1880 A more significant restriction is that jpeg_finish_compress() cannot suspend.
1881 This means you cannot use suspension with multi-pass operating modes, namely
1882 Huffman code optimization and multiple-scan output. Those modes write the
1883 whole file during jpeg_finish_compress(), which will certainly result in
1884 buffer overrun. (Note that this restriction applies only to compression,
1885 not decompression. The decompressor supports input suspension in all of its
1888 Decompression suspension:
1890 For decompression suspension, use a fill_input_buffer() routine that simply
1891 returns FALSE (except perhaps during error recovery, as discussed below).
1892 This will cause the decompressor to return to its caller with an indication
1893 that suspension has occurred. This can happen at four places:
1894 * jpeg_read_header(): will return JPEG_SUSPENDED.
1895 * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
1896 * jpeg*_read_scanlines(): will return the number of scanlines already
1897 completed (possibly 0).
1898 * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
1899 The surrounding application must recognize these cases, load more data into
1900 the input buffer, and repeat the call. In the case of jpeg*_read_scanlines(),
1901 increment the passed pointers past any scanlines successfully read.
1903 Just as with compression, the decompressor will typically backtrack to a
1904 convenient restart point before suspending. When fill_input_buffer() is
1905 called, next_input_byte/bytes_in_buffer point to the current restart point,
1906 which is where the decompressor will backtrack to if FALSE is returned.
1907 The data beyond that position must NOT be discarded if you suspend; it needs
1908 to be re-read upon resumption. In most implementations, you'll need to shift
1909 this data down to the start of your work buffer and then load more data after
1910 it. Again, this behavior means that a several-Kbyte work buffer is essential
1911 for decent performance; furthermore, you should load a reasonable amount of
1912 new data before resuming decompression. (If you loaded, say, only one new
1913 byte each time around, you could waste a LOT of cycles.)
1915 The skip_input_data() source manager routine requires special care in a
1916 suspension scenario. This routine is NOT granted the ability to suspend the
1917 decompressor; it can decrement bytes_in_buffer to zero, but no more. If the
1918 requested skip distance exceeds the amount of data currently in the input
1919 buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
1920 additional skip distance somewhere else. The decompressor will immediately
1921 call fill_input_buffer(), which should return FALSE, which will cause a
1922 suspension return. The surrounding application must then arrange to discard
1923 the recorded number of bytes before it resumes loading the input buffer.
1924 (Yes, this design is rather baroque, but it avoids complexity in the far more
1925 common case where a non-suspending source manager is used.)
1927 If the input data has been exhausted, we recommend that you emit a warning
1928 and insert dummy EOI markers just as a non-suspending data source manager
1929 would do. This can be handled either in the surrounding application logic or
1930 within fill_input_buffer(); the latter is probably more efficient. If
1931 fill_input_buffer() knows that no more data is available, it can set the
1932 pointer/count to point to a dummy EOI marker and then return TRUE just as
1933 though it had read more data in a non-suspending situation.
1935 The decompressor does not attempt to suspend within standard JPEG markers;
1936 instead it will backtrack to the start of the marker and reprocess the whole
1937 marker next time. Hence the input buffer must be large enough to hold the
1938 longest standard marker in the file. Standard JPEG markers should normally
1939 not exceed a few hundred bytes each (DHT tables are typically the longest).
1940 We recommend at least a 2K buffer for performance reasons, which is much
1941 larger than any correct marker is likely to be. For robustness against
1942 damaged marker length counts, you may wish to insert a test in your
1943 application for the case that the input buffer is completely full and yet
1944 the decoder has suspended without consuming any data --- otherwise, if this
1945 situation did occur, it would lead to an endless loop. (The library can't
1946 provide this test since it has no idea whether "the buffer is full", or
1947 even whether there is a fixed-size input buffer.)
1949 The input buffer would need to be 64K to allow for arbitrary COM or APPn
1950 markers, but these are handled specially: they are either saved into allocated
1951 memory, or skipped over by calling skip_input_data(). In the former case,
1952 suspension is handled correctly, and in the latter case, the problem of
1953 buffer overrun is placed on skip_input_data's shoulders, as explained above.
1954 Note that if you provide your own marker handling routine for large markers,
1955 you should consider how to deal with buffer overflow.
1957 Multiple-buffer management:
1959 In some applications it is desirable to store the compressed data in a linked
1960 list of buffer areas, so as to avoid data copying. This can be handled by
1961 having empty_output_buffer() or fill_input_buffer() set the pointer and count
1962 to reference the next available buffer; FALSE is returned only if no more
1963 buffers are available. Although seemingly straightforward, there is a
1964 pitfall in this approach: the backtrack that occurs when FALSE is returned
1965 could back up into an earlier buffer. For example, when fill_input_buffer()
1966 is called, the current pointer & count indicate the backtrack restart point.
1967 Since fill_input_buffer() will set the pointer and count to refer to a new
1968 buffer, the restart position must be saved somewhere else. Suppose a second
1969 call to fill_input_buffer() occurs in the same library call, and no
1970 additional input data is available, so fill_input_buffer must return FALSE.
1971 If the JPEG library has not moved the pointer/count forward in the current
1972 buffer, then *the correct restart point is the saved position in the prior
1973 buffer*. Prior buffers may be discarded only after the library establishes
1974 a restart point within a later buffer. Similar remarks apply for output into
1977 The library will never attempt to backtrack over a skip_input_data() call,
1978 so any skipped data can be permanently discarded. You still have to deal
1979 with the case of skipping not-yet-received data, however.
1981 It's much simpler to use only a single buffer; when fill_input_buffer() is
1982 called, move any unconsumed data (beyond the current pointer/count) down to
1983 the beginning of this buffer and then load new data into the remaining buffer
1984 space. This approach requires a little more data copying but is far easier
1988 Progressive JPEG support
1989 ------------------------
1991 Progressive JPEG rearranges the stored data into a series of scans of
1992 increasing quality. In situations where a JPEG file is transmitted across a
1993 slow communications link, a decoder can generate a low-quality image very
1994 quickly from the first scan, then gradually improve the displayed quality as
1995 more scans are received. The final image after all scans are complete is
1996 identical to that of a regular (sequential) JPEG file of the same quality
1997 setting. Progressive JPEG files are often slightly smaller than equivalent
1998 sequential JPEG files, but the possibility of incremental display is the main
1999 reason for using progressive JPEG.
2001 The IJG encoder library generates progressive JPEG files when given a
2002 suitable "scan script" defining how to divide the data into scans.
2003 Creation of progressive JPEG files is otherwise transparent to the encoder.
2004 Progressive JPEG files can also be read transparently by the decoder library.
2005 If the decoding application simply uses the library as defined above, it
2006 will receive a final decoded image without any indication that the file was
2007 progressive. Of course, this approach does not allow incremental display.
2008 To perform incremental display, an application needs to use the decoder
2009 library's "buffered-image" mode, in which it receives a decoded image
2012 Each displayed scan requires about as much work to decode as a full JPEG
2013 image of the same size, so the decoder must be fairly fast in relation to the
2014 data transmission rate in order to make incremental display useful. However,
2015 it is possible to skip displaying the image and simply add the incoming bits
2016 to the decoder's coefficient buffer. This is fast because only Huffman
2017 decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
2018 The IJG decoder library allows the application to switch dynamically between
2019 displaying the image and simply absorbing the incoming bits. A properly
2020 coded application can automatically adapt the number of display passes to
2021 suit the time available as the image is received. Also, a final
2022 higher-quality display cycle can be performed from the buffered data after
2023 the end of the file is reached.
2025 Progressive compression:
2027 To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
2028 set the scan_info cinfo field to point to an array of scan descriptors, and
2029 perform compression as usual. Instead of constructing your own scan list,
2030 you can call the jpeg_simple_progression() helper routine to create a
2031 recommended progression sequence; this method should be used by all
2032 applications that don't want to get involved in the nitty-gritty of
2033 progressive scan sequence design. (If you want to provide user control of
2034 scan sequences, you may wish to borrow the scan script reading code found
2035 in rdswitch.c, so that you can read scan script files just like cjpeg's.)
2036 When scan_info is not NULL, the compression library will store DCT'd data
2037 into a buffer array as jpeg*_write_scanlines() is called, and will emit all
2038 the requested scans during jpeg_finish_compress(). This implies that
2039 multiple-scan output cannot be created with a suspending data destination
2040 manager, since jpeg_finish_compress() does not support suspension. We
2041 should also note that the compressor currently forces Huffman optimization
2042 mode when creating a progressive JPEG file, because the default Huffman
2043 tables are unsuitable for progressive files.
2045 Progressive decompression:
2047 When buffered-image mode is not used, the decoder library will read all of
2048 a multi-scan file during jpeg_start_decompress(), so that it can provide a
2049 final decoded image. (Here "multi-scan" means either progressive or
2050 multi-scan sequential.) This makes multi-scan files transparent to the
2051 decoding application. However, existing applications that used suspending
2052 input with version 5 of the IJG library will need to be modified to check
2053 for a suspension return from jpeg_start_decompress().
2055 To perform incremental display, an application must use the library's
2056 buffered-image mode. This is described in the next section.
2062 In buffered-image mode, the library stores the partially decoded image in a
2063 coefficient buffer, from which it can be read out as many times as desired.
2064 This mode is typically used for incremental display of progressive JPEG files,
2065 but it can be used with any JPEG file. Each scan of a progressive JPEG file
2066 adds more data (more detail) to the buffered image. The application can
2067 display in lockstep with the source file (one display pass per input scan),
2068 or it can allow input processing to outrun display processing. By making
2069 input and display processing run independently, it is possible for the
2070 application to adapt progressive display to a wide range of data transmission
2073 The basic control flow for buffered-image decoding is
2075 jpeg_create_decompress()
2078 set overall decompression parameters
2079 cinfo.buffered_image = TRUE; /* select buffered-image mode */
2080 jpeg_start_decompress()
2081 for (each output pass) {
2082 adjust output decompression parameters if required
2083 jpeg_start_output() /* start a new output pass */
2084 for (all scanlines in image) {
2085 jpeg_read_scanlines() /* Use jpeg12_read_scanlines() for
2086 12-bit data precision and
2087 jpeg16_read_scanlines() for 16-bit
2091 jpeg_finish_output() /* terminate output pass */
2093 jpeg_finish_decompress()
2094 jpeg_destroy_decompress()
2096 This differs from ordinary unbuffered decoding in that there is an additional
2097 level of looping. The application can choose how many output passes to make
2098 and how to display each pass.
2100 The simplest approach to displaying progressive images is to do one display
2101 pass for each scan appearing in the input file. In this case the outer loop
2102 condition is typically
2103 while (!jpeg_input_complete(&cinfo))
2104 and the start-output call should read
2105 jpeg_start_output(&cinfo, cinfo.input_scan_number);
2106 The second parameter to jpeg_start_output() indicates which scan of the input
2107 file is to be displayed; the scans are numbered starting at 1 for this
2108 purpose. (You can use a loop counter starting at 1 if you like, but using
2109 the library's input scan counter is easier.) The library automatically reads
2110 data as necessary to complete each requested scan, and jpeg_finish_output()
2111 advances to the next scan or end-of-image marker (hence input_scan_number
2112 will be incremented by the time control arrives back at jpeg_start_output()).
2113 With this technique, data is read from the input file only as needed, and
2114 input and output processing run in lockstep.
2116 After reading the final scan and reaching the end of the input file, the
2117 buffered image remains available; it can be read additional times by
2118 repeating the jpeg_start_output()/jpeg*_read_scanlines()/jpeg_finish_output()
2119 sequence. For example, a useful technique is to use fast one-pass color
2120 quantization for display passes made while the image is arriving, followed by
2121 a final display pass using two-pass quantization for highest quality. This
2122 is done by changing the library parameters before the final output pass.
2123 Changing parameters between passes is discussed in detail below.
2125 In general the last scan of a progressive file cannot be recognized as such
2126 until after it is read, so a post-input display pass is the best approach if
2127 you want special processing in the final pass.
2129 When done with the image, be sure to call jpeg_finish_decompress() to release
2130 the buffered image (or just use jpeg_destroy_decompress()).
2132 If input data arrives faster than it can be displayed, the application can
2133 cause the library to decode input data in advance of what's needed to produce
2134 output. This is done by calling the routine jpeg_consume_input().
2135 The return value is one of the following:
2136 JPEG_REACHED_SOS: reached an SOS marker (the start of a new scan)
2137 JPEG_REACHED_EOI: reached the EOI marker (end of image)
2138 JPEG_ROW_COMPLETED: completed reading one MCU row of compressed data
2139 JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
2140 JPEG_SUSPENDED: suspended before completing any of the above
2141 (JPEG_SUSPENDED can occur only if a suspending data source is used.) This
2142 routine can be called at any time after initializing the JPEG object. It
2143 reads some additional data and returns when one of the indicated significant
2144 events occurs. (If called after the EOI marker is reached, it will
2145 immediately return JPEG_REACHED_EOI without attempting to read more data.)
2147 The library's output processing will automatically call jpeg_consume_input()
2148 whenever the output processing overtakes the input; thus, simple lockstep
2149 display requires no direct calls to jpeg_consume_input(). But by adding
2150 calls to jpeg_consume_input(), you can absorb data in advance of what is
2151 being displayed. This has two benefits:
2152 * You can limit buildup of unprocessed data in your input buffer.
2153 * You can eliminate extra display passes by paying attention to the
2154 state of the library's input processing.
2156 The first of these benefits only requires interspersing calls to
2157 jpeg_consume_input() with your display operations and any other processing
2158 you may be doing. To avoid wasting cycles due to backtracking, it's best to
2159 call jpeg_consume_input() only after a hundred or so new bytes have arrived.
2160 This is discussed further under "I/O suspension", above. (Note: the JPEG
2161 library currently is not thread-safe. You must not call jpeg_consume_input()
2162 from one thread of control if a different library routine is working on the
2163 same JPEG object in another thread.)
2165 When input arrives fast enough that more than one new scan is available
2166 before you start a new output pass, you may as well skip the output pass
2167 corresponding to the completed scan. This occurs for free if you pass
2168 cinfo.input_scan_number as the target scan number to jpeg_start_output().
2169 The input_scan_number field is simply the index of the scan currently being
2170 consumed by the input processor. You can ensure that this is up-to-date by
2171 emptying the input buffer just before calling jpeg_start_output(): call
2172 jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
2175 The target scan number passed to jpeg_start_output() is saved in the
2176 cinfo.output_scan_number field. The library's output processing calls
2177 jpeg_consume_input() whenever the current input scan number and row within
2178 that scan is less than or equal to the current output scan number and row.
2179 Thus, input processing can "get ahead" of the output processing but is not
2180 allowed to "fall behind". You can achieve several different effects by
2181 manipulating this interlock rule. For example, if you pass a target scan
2182 number greater than the current input scan number, the output processor will
2183 wait until that scan starts to arrive before producing any output. (To avoid
2184 an infinite loop, the target scan number is automatically reset to the last
2185 scan number when the end of image is reached. Thus, if you specify a large
2186 target scan number, the library will just absorb the entire input file and
2187 then perform an output pass. This is effectively the same as what
2188 jpeg_start_decompress() does when you don't select buffered-image mode.)
2189 When you pass a target scan number equal to the current input scan number,
2190 the image is displayed no faster than the current input scan arrives. The
2191 final possibility is to pass a target scan number less than the current input
2192 scan number; this disables the input/output interlock and causes the output
2193 processor to simply display whatever it finds in the image buffer, without
2194 waiting for input. (However, the library will not accept a target scan
2195 number less than one, so you can't avoid waiting for the first scan.)
2197 When data is arriving faster than the output display processing can advance
2198 through the image, jpeg_consume_input() will store data into the buffered
2199 image beyond the point at which the output processing is reading data out
2200 again. If the input arrives fast enough, it may "wrap around" the buffer to
2201 the point where the input is more than one whole scan ahead of the output.
2202 If the output processing simply proceeds through its display pass without
2203 paying attention to the input, the effect seen on-screen is that the lower
2204 part of the image is one or more scans better in quality than the upper part.
2205 Then, when the next output scan is started, you have a choice of what target
2206 scan number to use. The recommended choice is to use the current input scan
2207 number at that time, which implies that you've skipped the output scans
2208 corresponding to the input scans that were completed while you processed the
2209 previous output scan. In this way, the decoder automatically adapts its
2210 speed to the arriving data, by skipping output scans as necessary to keep up
2211 with the arriving data.
2213 When using this strategy, you'll want to be sure that you perform a final
2214 output pass after receiving all the data; otherwise your last display may not
2215 be full quality across the whole screen. So the right outer loop logic is
2216 something like this:
2218 absorb any waiting input by calling jpeg_consume_input()
2219 final_pass = jpeg_input_complete(&cinfo);
2220 adjust output decompression parameters if required
2221 jpeg_start_output(&cinfo, cinfo.input_scan_number);
2223 jpeg_finish_output()
2224 } while (!final_pass);
2225 rather than quitting as soon as jpeg_input_complete() returns TRUE. This
2226 arrangement makes it simple to use higher-quality decoding parameters
2227 for the final pass. But if you don't want to use special parameters for
2228 the final pass, the right loop logic is like this:
2230 absorb any waiting input by calling jpeg_consume_input()
2231 jpeg_start_output(&cinfo, cinfo.input_scan_number);
2233 jpeg_finish_output()
2234 if (jpeg_input_complete(&cinfo) &&
2235 cinfo.input_scan_number == cinfo.output_scan_number)
2238 In this case you don't need to know in advance whether an output pass is to
2239 be the last one, so it's not necessary to have reached EOF before starting
2240 the final output pass; rather, what you want to test is whether the output
2241 pass was performed in sync with the final input scan. This form of the loop
2242 will avoid an extra output pass whenever the decoder is able (or nearly able)
2243 to keep up with the incoming data.
2245 When the data transmission speed is high, you might begin a display pass,
2246 then find that much or all of the file has arrived before you can complete
2247 the pass. (You can detect this by noting the JPEG_REACHED_EOI return code
2248 from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
2249 In this situation you may wish to abort the current display pass and start a
2250 new one using the newly arrived information. To do so, just call
2251 jpeg_finish_output() and then start a new pass with jpeg_start_output().
2253 A variant strategy is to abort and restart display if more than one complete
2254 scan arrives during an output pass; this can be detected by noting
2255 JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number. This
2256 idea should be employed with caution, however, since the display process
2257 might never get to the bottom of the image before being aborted, resulting
2258 in the lower part of the screen being several passes worse than the upper.
2259 In most cases it's probably best to abort an output pass only if the whole
2260 file has arrived and you want to begin the final output pass immediately.
2262 When receiving data across a communication link, we recommend always using
2263 the current input scan number for the output target scan number; if a
2264 higher-quality final pass is to be done, it should be started (aborting any
2265 incomplete output pass) as soon as the end of file is received. However,
2266 many other strategies are possible. For example, the application can examine
2267 the parameters of the current input scan and decide whether to display it or
2268 not. If the scan contains only chroma data, one might choose not to use it
2269 as the target scan, expecting that the scan will be small and will arrive
2270 quickly. To skip to the next scan, call jpeg_consume_input() until it
2271 returns JPEG_REACHED_SOS or JPEG_REACHED_EOI. Or just use the next higher
2272 number as the target scan for jpeg_start_output(); but that method doesn't
2273 let you inspect the next scan's parameters before deciding to display it.
2276 In buffered-image mode, jpeg_start_decompress() never performs input and
2277 thus never suspends. An application that uses input suspension with
2278 buffered-image mode must be prepared for suspension returns from these
2280 * jpeg_start_output() performs input only if you request 2-pass quantization
2281 and the target scan isn't fully read yet. (This is discussed below.)
2282 * jpeg*_read_scanlines(), as always, returns the number of scanlines that it
2283 was able to produce before suspending.
2284 * jpeg_finish_output() will read any markers following the target scan,
2285 up to the end of the file or the SOS marker that begins another scan.
2286 (But it reads no input if jpeg_consume_input() has already reached the
2287 end of the file or a SOS marker beyond the target output scan.)
2288 * jpeg_finish_decompress() will read until the end of file, and thus can
2289 suspend if the end hasn't already been reached (as can be tested by
2290 calling jpeg_input_complete()).
2291 jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
2292 all return TRUE if they completed their tasks, FALSE if they had to suspend.
2293 In the event of a FALSE return, the application must load more input data
2294 and repeat the call. Applications that use non-suspending data sources need
2295 not check the return values of these three routines.
2298 It is possible to change decoding parameters between output passes in the
2299 buffered-image mode. The decoder library currently supports only very
2300 limited changes of parameters. ONLY THE FOLLOWING parameter changes are
2301 allowed after jpeg_start_decompress() is called:
2302 * dct_method can be changed before each call to jpeg_start_output().
2303 For example, one could use a fast DCT method for early scans, changing
2304 to a higher quality method for the final scan.
2305 * dither_mode can be changed before each call to jpeg_start_output();
2306 of course this has no impact if not using color quantization. Typically
2307 one would use ordered dither for initial passes, then switch to
2308 Floyd-Steinberg dither for the final pass. Caution: changing dither mode
2309 can cause more memory to be allocated by the library. Although the amount
2310 of memory involved is not large (a scanline or so), it may cause the
2311 initial max_memory_to_use specification to be exceeded, which in the worst
2312 case would result in an out-of-memory failure.
2313 * do_block_smoothing can be changed before each call to jpeg_start_output().
2314 This setting is relevant only when decoding a progressive JPEG image.
2315 During the first DC-only scan, block smoothing provides a very "fuzzy" look
2316 instead of the very "blocky" look seen without it; which is better seems a
2317 matter of personal taste. But block smoothing is nearly always a win
2318 during later stages, especially when decoding a successive-approximation
2319 image: smoothing helps to hide the slight blockiness that otherwise shows
2320 up on smooth gradients until the lowest coefficient bits are sent.
2321 * Color quantization mode can be changed under the rules described below.
2322 You *cannot* change between full-color and quantized output (because that
2323 would alter the required I/O buffer sizes), but you can change which
2324 quantization method is used.
2326 When generating color-quantized output, changing quantization method is a
2327 very useful way of switching between high-speed and high-quality display.
2328 The library allows you to change among its three quantization methods:
2329 1. Single-pass quantization to a fixed color cube.
2330 Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
2331 2. Single-pass quantization to an application-supplied colormap.
2332 Selected by setting cinfo.colormap to point to the colormap (the value of
2333 two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
2334 3. Two-pass quantization to a colormap chosen specifically for the image.
2335 Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
2336 (This is the default setting selected by jpeg_read_header, but it is
2337 probably NOT what you want for the first pass of progressive display!)
2338 These methods offer successively better quality and lesser speed. However,
2339 only the first method is available for quantizing in non-RGB color spaces.
2341 IMPORTANT: because the different quantizer methods have very different
2342 working-storage requirements, the library requires you to indicate which
2343 one(s) you intend to use before you call jpeg_start_decompress(). (If we did
2344 not require this, the max_memory_to_use setting would be a complete fiction.)
2345 You do this by setting one or more of these three cinfo fields to TRUE:
2346 enable_1pass_quant Fixed color cube colormap
2347 enable_external_quant Externally-supplied colormap
2348 enable_2pass_quant Two-pass custom colormap
2349 All three are initialized FALSE by jpeg_read_header(). But
2350 jpeg_start_decompress() automatically sets TRUE the one selected by the
2351 current two_pass_quantize and colormap settings, so you only need to set the
2352 enable flags for any other quantization methods you plan to change to later.
2354 After setting the enable flags correctly at jpeg_start_decompress() time, you
2355 can change to any enabled quantization method by setting two_pass_quantize
2356 and colormap properly just before calling jpeg_start_output(). The following
2357 special rules apply:
2358 1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
2359 or 2-pass mode from a different mode, or when you want the 2-pass
2360 quantizer to be re-run to generate a new colormap.
2361 2. To switch to an external colormap, or to change to a different external
2362 colormap than was used on the prior pass, you must call
2363 jpeg_new_colormap() after setting cinfo.colormap.
2364 NOTE: if you want to use the same colormap as was used in the prior pass,
2365 you should not do either of these things. This will save some nontrivial
2367 (These requirements exist because cinfo.colormap will always be non-NULL
2368 after completing a prior output pass, since both the 1-pass and 2-pass
2369 quantizers set it to point to their output colormaps. Thus you have to
2370 do one of these two things to notify the library that something has changed.
2371 Yup, it's a bit klugy, but it's necessary to do it this way for backwards
2374 Note that in buffered-image mode, the library generates any requested colormap
2375 during jpeg_start_output(), not during jpeg_start_decompress().
2377 When using two-pass quantization, jpeg_start_output() makes a pass over the
2378 buffered image to determine the optimum color map; it therefore may take a
2379 significant amount of time, whereas ordinarily it does little work. The
2380 progress monitor hook is called during this pass, if defined. It is also
2381 important to realize that if the specified target scan number is greater than
2382 or equal to the current input scan number, jpeg_start_output() will attempt
2383 to consume input as it makes this pass. If you use a suspending data source,
2384 you need to check for a FALSE return from jpeg_start_output() under these
2385 conditions. The combination of 2-pass quantization and a not-yet-fully-read
2386 target scan is the only case in which jpeg_start_output() will consume input.
2389 Application authors who support buffered-image mode may be tempted to use it
2390 for all JPEG images, even single-scan ones. This will work, but it is
2391 inefficient: there is no need to create an image-sized coefficient buffer for
2392 single-scan images. Requesting buffered-image mode for such an image wastes
2393 memory. Worse, it can cost time on large images, since the buffered data has
2394 to be swapped out or written to a temporary file. If you are concerned about
2395 maximum performance on baseline JPEG files, you should use buffered-image
2396 mode only when the incoming file actually has multiple scans. This can be
2397 tested by calling jpeg_has_multiple_scans(), which will return a correct
2398 result at any time after jpeg_read_header() completes.
2400 It is also worth noting that when you use jpeg_consume_input() to let input
2401 processing get ahead of output processing, the resulting pattern of access to
2402 the coefficient buffer is quite nonsequential. It's best to use the memory
2403 manager jmemnobs.c if you can (ie, if you have enough real or virtual main
2404 memory). If not, at least make sure that max_memory_to_use is set as high as
2405 possible. If the JPEG memory manager has to use a temporary file, you will
2406 probably see a lot of disk traffic and poor performance. (This could be
2407 improved with additional work on the memory manager, but we haven't gotten
2410 In some applications it may be convenient to use jpeg_consume_input() for all
2411 input processing, including reading the initial markers; that is, you may
2412 wish to call jpeg_consume_input() instead of jpeg_read_header() during
2413 startup. This works, but note that you must check for JPEG_REACHED_SOS and
2414 JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
2415 Once the first SOS marker has been reached, you must call
2416 jpeg_start_decompress() before jpeg_consume_input() will consume more input;
2417 it'll just keep returning JPEG_REACHED_SOS until you do. If you read a
2418 tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
2419 without ever returning JPEG_REACHED_SOS; be sure to check for this case.
2420 If this happens, the decompressor will not read any more input until you call
2421 jpeg_abort() to reset it. It is OK to call jpeg_consume_input() even when not
2422 using buffered-image mode, but in that case it's basically a no-op after the
2423 initial markers have been read: it will just return JPEG_SUSPENDED.
2426 Abbreviated datastreams and multiple images
2427 -------------------------------------------
2429 A JPEG compression or decompression object can be reused to process multiple
2430 images. This saves a small amount of time per image by eliminating the
2431 "create" and "destroy" operations, but that isn't the real purpose of the
2432 feature. Rather, reuse of an object provides support for abbreviated JPEG
2433 datastreams. Object reuse can also simplify processing a series of images in
2434 a single input or output file. This section explains these features.
2436 A JPEG file normally contains several hundred bytes worth of quantization
2437 and Huffman tables. In a situation where many images will be stored or
2438 transmitted with identical tables, this may represent an annoying overhead.
2439 The JPEG standard therefore permits tables to be omitted. The standard
2440 defines three classes of JPEG datastreams:
2441 * "Interchange" datastreams contain an image and all tables needed to decode
2442 the image. These are the usual kind of JPEG file.
2443 * "Abbreviated image" datastreams contain an image, but are missing some or
2444 all of the tables needed to decode that image.
2445 * "Abbreviated table specification" (henceforth "tables-only") datastreams
2446 contain only table specifications.
2447 To decode an abbreviated image, it is necessary to load the missing table(s)
2448 into the decoder beforehand. This can be accomplished by reading a separate
2449 tables-only file. A variant scheme uses a series of images in which the first
2450 image is an interchange (complete) datastream, while subsequent ones are
2451 abbreviated and rely on the tables loaded by the first image. It is assumed
2452 that once the decoder has read a table, it will remember that table until a
2453 new definition for the same table number is encountered.
2455 It is the application designer's responsibility to figure out how to associate
2456 the correct tables with an abbreviated image. While abbreviated datastreams
2457 can be useful in a closed environment, their use is strongly discouraged in
2458 any situation where data exchange with other applications might be needed.
2461 The JPEG library provides support for reading and writing any combination of
2462 tables-only datastreams and abbreviated images. In both compression and
2463 decompression objects, a quantization or Huffman table will be retained for
2464 the lifetime of the object, unless it is overwritten by a new table definition.
2467 To create abbreviated image datastreams, it is only necessary to tell the
2468 compressor not to emit some or all of the tables it is using. Each
2469 quantization and Huffman table struct contains a boolean field "sent_table",
2470 which normally is initialized to FALSE. For each table used by the image, the
2471 header-writing process emits the table and sets sent_table = TRUE unless it is
2472 already TRUE. (In normal usage, this prevents outputting the same table
2473 definition multiple times, as would otherwise occur because the chroma
2474 components typically share tables.) Thus, setting this field to TRUE before
2475 calling jpeg_start_compress() will prevent the table from being written at
2478 If you want to create a "pure" abbreviated image file containing no tables,
2479 just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
2480 tables. If you want to emit some but not all tables, you'll need to set the
2481 individual sent_table fields directly.
2483 To create an abbreviated image, you must also call jpeg_start_compress()
2484 with a second parameter of FALSE, not TRUE. Otherwise jpeg_start_compress()
2485 will force all the sent_table fields to FALSE. (This is a safety feature to
2486 prevent abbreviated images from being created accidentally.)
2488 To create a tables-only file, perform the same parameter setup that you
2489 normally would, but instead of calling jpeg_start_compress() and so on, call
2490 jpeg_write_tables(&cinfo). This will write an abbreviated datastream
2491 containing only SOI, DQT and/or DHT markers, and EOI. All the quantization
2492 and Huffman tables that are currently defined in the compression object will
2493 be emitted unless their sent_tables flag is already TRUE, and then all the
2494 sent_tables flags will be set TRUE.
2496 A sure-fire way to create matching tables-only and abbreviated image files
2497 is to proceed as follows:
2499 create JPEG compression object
2501 set destination to tables-only file
2502 jpeg_write_tables(&cinfo);
2503 set destination to image file
2504 jpeg_start_compress(&cinfo, FALSE);
2506 jpeg_finish_compress(&cinfo);
2508 Since the JPEG parameters are not altered between writing the table file and
2509 the abbreviated image file, the same tables are sure to be used. Of course,
2510 you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
2511 many times to produce many abbreviated image files matching the table file.
2513 You cannot suppress output of the computed Huffman tables when Huffman
2514 optimization is selected. (If you could, there'd be no way to decode the
2515 image...) Generally, you don't want to set optimize_coding = TRUE when
2516 you are trying to produce abbreviated files.
2518 In some cases you might want to compress an image using tables which are
2519 not stored in the application, but are defined in an interchange or
2520 tables-only file readable by the application. This can be done by setting up
2521 a JPEG decompression object to read the specification file, then copying the
2522 tables into your compression object. See jpeg_copy_critical_parameters()
2523 for an example of copying quantization tables.
2526 To read abbreviated image files, you simply need to load the proper tables
2527 into the decompression object before trying to read the abbreviated image.
2528 If the proper tables are stored in the application program, you can just
2529 allocate the table structs and fill in their contents directly. For example,
2530 to load a fixed quantization table into table slot "n":
2532 if (cinfo.quant_tbl_ptrs[n] == NULL)
2533 cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
2534 quant_ptr = cinfo.quant_tbl_ptrs[n]; /* quant_ptr is JQUANT_TBL* */
2535 for (i = 0; i < 64; i++) {
2536 /* Qtable[] is desired quantization table, in natural array order */
2537 quant_ptr->quantval[i] = Qtable[i];
2540 Code to load a fixed Huffman table is typically (for AC table "n"):
2542 if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
2543 cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
2544 huff_ptr = cinfo.ac_huff_tbl_ptrs[n]; /* huff_ptr is JHUFF_TBL* */
2545 for (i = 1; i <= 16; i++) {
2546 /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
2547 huff_ptr->bits[i] = counts[i];
2549 for (i = 0; i < 256; i++) {
2550 /* symbols[] is the list of Huffman symbols, in code-length order */
2551 huff_ptr->huffval[i] = symbols[i];
2554 (Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
2555 constant JQUANT_TBL object is not safe. If the incoming file happened to
2556 contain a quantization table definition, your master table would get
2557 overwritten! Instead allocate a working table copy and copy the master table
2558 into it, as illustrated above. Ditto for Huffman tables, of course.)
2560 You might want to read the tables from a tables-only file, rather than
2561 hard-wiring them into your application. The jpeg_read_header() call is
2562 sufficient to read a tables-only file. You must pass a second parameter of
2563 FALSE to indicate that you do not require an image to be present. Thus, the
2566 create JPEG decompression object
2567 set source to tables-only file
2568 jpeg_read_header(&cinfo, FALSE);
2569 set source to abbreviated image file
2570 jpeg_read_header(&cinfo, TRUE);
2571 set decompression parameters
2572 jpeg_start_decompress(&cinfo);
2574 jpeg_finish_decompress(&cinfo);
2576 In some cases, you may want to read a file without knowing whether it contains
2577 an image or just tables. In that case, pass FALSE and check the return value
2578 from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
2579 JPEG_HEADER_TABLES_ONLY if only tables were found. (A third return value,
2580 JPEG_SUSPENDED, is possible when using a suspending data source manager.)
2581 Note that jpeg_read_header() will not complain if you read an abbreviated
2582 image for which you haven't loaded the missing tables; the missing-table check
2583 occurs later, in jpeg_start_decompress().
2586 It is possible to read a series of images from a single source file by
2587 repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
2588 without releasing/recreating the JPEG object or the data source module.
2589 (If you did reinitialize, any partial bufferload left in the data source
2590 buffer at the end of one image would be discarded, causing you to lose the
2591 start of the next image.) When you use this method, stored tables are
2592 automatically carried forward, so some of the images can be abbreviated images
2593 that depend on tables from earlier images.
2595 If you intend to write a series of images into a single destination file,
2596 you might want to make a specialized data destination module that doesn't
2597 flush the output buffer at term_destination() time. This would speed things
2598 up by some trifling amount. Of course, you'd need to remember to flush the
2599 buffer after the last image. You can make the later images be abbreviated
2600 ones by passing FALSE to jpeg_start_compress().
2606 Some applications may need to insert or extract special data in the JPEG
2607 datastream. The JPEG standard provides marker types "COM" (comment) and
2608 "APP0" through "APP15" (application) to hold application-specific data.
2609 Unfortunately, the use of these markers is not specified by the standard.
2610 COM markers are fairly widely used to hold user-supplied text. The JFIF file
2611 format spec uses APP0 markers with specified initial strings to hold certain
2612 data. Adobe applications use APP14 markers beginning with the string "Adobe"
2613 for miscellaneous data. Other APPn markers are rarely seen, but might
2614 contain almost anything.
2616 If you wish to store user-supplied text, we recommend you use COM markers
2617 and place readable 7-bit ASCII text in them. Newline conventions are not
2618 standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
2619 (Mac style). A robust COM reader should be able to cope with random binary
2620 garbage, including nulls, since some applications generate COM markers
2621 containing non-ASCII junk. (But yours should not be one of them.)
2623 For program-supplied data, use an APPn marker, and be sure to begin it with an
2624 identifying string so that you can tell whether the marker is actually yours.
2625 It's probably best to avoid using APP0 or APP14 for any private markers.
2626 (NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
2627 not use APP8 markers for any private purposes, either.)
2629 Keep in mind that at most 65533 bytes can be put into one marker, but you
2630 can have as many markers as you like.
2632 By default, the IJG compression library will write a JFIF APP0 marker if the
2633 selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
2634 the selected colorspace is RGB, CMYK, or YCCK. You can disable this, but
2635 we don't recommend it. The decompression library will recognize JFIF and
2636 Adobe markers and will set the JPEG colorspace properly when one is found.
2639 You can write special markers immediately following the datastream header by
2640 calling jpeg_write_marker() after jpeg_start_compress() and before the first
2641 call to jpeg*_write_scanlines(). When you do this, the markers appear after
2642 the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
2643 all else. Specify the marker type parameter as "JPEG_COM" for COM or
2644 "JPEG_APP0 + n" for APPn. (Actually, jpeg_write_marker will let you write
2645 any marker type, but we don't recommend writing any other kinds of marker.)
2646 For example, to write a user comment string pointed to by comment_text:
2647 jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
2649 If it's not convenient to store all the marker data in memory at once,
2650 you can instead call jpeg_write_m_header() followed by multiple calls to
2651 jpeg_write_m_byte(). If you do it this way, it's your responsibility to
2652 call jpeg_write_m_byte() exactly the number of times given in the length
2653 parameter to jpeg_write_m_header(). (This method lets you empty the
2654 output buffer partway through a marker, which might be important when
2655 using a suspending data destination module. In any case, if you are using
2656 a suspending destination, you should flush its buffer after inserting
2657 any special markers. See "I/O suspension".)
2659 Or, if you prefer to synthesize the marker byte sequence yourself,
2660 you can just cram it straight into the data destination module.
2662 If you are writing JFIF 1.02 extension markers (thumbnail images), don't
2663 forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
2664 correct JFIF version number in the JFIF header marker. The library's default
2665 is to write version 1.01, but that's wrong if you insert any 1.02 extension
2666 markers. (We could probably get away with just defaulting to 1.02, but there
2667 used to be broken decoders that would complain about unknown minor version
2668 numbers. To reduce compatibility risks it's safest not to write 1.02 unless
2669 you are actually using 1.02 extensions.)
2672 When reading, two methods of handling special markers are available:
2673 1. You can ask the library to save the contents of COM and/or APPn markers
2674 into memory, and then examine them at your leisure afterwards.
2675 2. You can supply your own routine to process COM and/or APPn markers
2676 on-the-fly as they are read.
2677 The first method is simpler to use, especially if you are using a suspending
2678 data source; writing a marker processor that copes with input suspension is
2679 not easy (consider what happens if the marker is longer than your available
2680 input buffer). However, the second method conserves memory since the marker
2681 data need not be kept around after it's been processed.
2683 For either method, you'd normally set up marker handling after creating a
2684 decompression object and before calling jpeg_read_header(), because the
2685 markers of interest will typically be near the head of the file and so will
2686 be scanned by jpeg_read_header. Once you've established a marker handling
2687 method, it will be used for the life of that decompression object
2688 (potentially many datastreams), unless you change it. Marker handling is
2689 determined separately for COM markers and for each APPn marker code.
2692 To save the contents of special markers in memory, call
2693 jpeg_save_markers(cinfo, marker_code, length_limit)
2694 where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
2695 (To arrange to save all the special marker types, you need to call this
2696 routine 17 times, for COM and APP0-APP15.) If the incoming marker is longer
2697 than length_limit data bytes, only length_limit bytes will be saved; this
2698 parameter allows you to avoid chewing up memory when you only need to see the
2699 first few bytes of a potentially large marker. If you want to save all the
2700 data, set length_limit to 0xFFFF; that is enough since marker lengths are only
2701 16 bits. As a special case, setting length_limit to 0 prevents that marker
2702 type from being saved at all. (That is the default behavior, in fact.)
2704 After jpeg_read_header() completes, you can examine the special markers by
2705 following the cinfo->marker_list pointer chain. All the special markers in
2706 the file appear in this list, in order of their occurrence in the file (but
2707 omitting any markers of types you didn't ask for). Both the original data
2708 length and the saved data length are recorded for each list entry; the latter
2709 will not exceed length_limit for the particular marker type. Note that these
2710 lengths exclude the marker length word, whereas the stored representation
2711 within the JPEG file includes it. (Hence the maximum data length is really
2714 It is possible that additional special markers appear in the file beyond the
2715 SOS marker at which jpeg_read_header stops; if so, the marker list will be
2716 extended during reading of the rest of the file. This is not expected to be
2717 common, however. If you are short on memory you may want to reset the length
2718 limit to zero for all marker types after finishing jpeg_read_header, to
2719 ensure that the max_memory_to_use setting cannot be exceeded due to addition
2722 The marker list remains stored until you call jpeg_finish_decompress or
2723 jpeg_abort, at which point the memory is freed and the list is set to empty.
2724 (jpeg_destroy also releases the storage, of course.)
2726 Note that the library is internally interested in APP0 and APP14 markers;
2727 if you try to set a small nonzero length limit on these types, the library
2728 will silently force the length up to the minimum it wants. (But you can set
2729 a zero length limit to prevent them from being saved at all.) Also, in a
2730 16-bit environment, the maximum length limit may be constrained to less than
2731 65533 by malloc() limitations. It is therefore best not to assume that the
2732 effective length limit is exactly what you set it to be.
2735 If you want to supply your own marker-reading routine, you do it by calling
2736 jpeg_set_marker_processor(). A marker processor routine must have the
2738 boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
2739 Although the marker code is not explicitly passed, the routine can find it
2740 in cinfo->unread_marker. At the time of call, the marker proper has been
2741 read from the data source module. The processor routine is responsible for
2742 reading the marker length word and the remaining parameter bytes, if any.
2743 Return TRUE to indicate success. (FALSE should be returned only if you are
2744 using a suspending data source and it tells you to suspend. See the standard
2745 marker processors in jdmarker.c for appropriate coding methods if you need to
2746 use a suspending data source.)
2748 If you override the default APP0 or APP14 processors, it is up to you to
2749 recognize JFIF and Adobe markers if you want colorspace recognition to occur
2750 properly. We recommend copying and extending the default processors if you
2751 want to do that. (A better idea is to save these marker types for later
2752 examination by calling jpeg_save_markers(); that method doesn't interfere
2753 with the library's own processing of these markers.)
2755 jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
2756 --- if you call one it overrides any previous call to the other, for the
2757 particular marker type specified.
2759 A simple example of an external COM processor can be found in djpeg.c.
2760 Also, see jpegtran.c for an example of using jpeg_save_markers.
2766 Two functions are provided for writing and reading International Color
2767 Consortium (ICC) device profiles embedded in JFIF JPEG image files:
2769 void jpeg_write_icc_profile (j_compress_ptr cinfo,
2770 const JOCTET *icc_data_ptr,
2771 unsigned int icc_data_len);
2772 boolean jpeg_read_icc_profile (j_decompress_ptr cinfo,
2773 JOCTET **icc_data_ptr,
2774 unsigned int *icc_data_len);
2776 The ICC has defined a standard for including such data in JPEG "APP2" markers.
2777 The aforementioned functions do not know anything about the internal structure
2778 of the ICC profile data; they just know how to embed the profile data into a
2779 JPEG file while writing it, or to extract the profile data from a JPEG file
2782 jpeg_write_icc_profile() must be called after calling jpeg_start_compress() and
2783 before the first call to jpeg*_write_scanlines() or jpeg*_write_raw_data().
2784 This ordering ensures that the APP2 marker(s) will appear after the SOI and
2785 JFIF or Adobe markers, but before all other data.
2787 jpeg_read_icc_profile() returns TRUE if an ICC profile was found and FALSE
2788 otherwise. If an ICC profile was found, then the function will allocate a
2789 memory region containing the profile and will return a pointer to that memory
2790 region in *icc_data_ptr, as well as the length of the region in *icc_data_len.
2791 This memory region is allocated by the library using malloc() and must be freed
2792 by the caller using free() when the memory region is no longer needed. Callers
2793 wishing to use jpeg_read_icc_profile() must call
2795 jpeg_save_markers(cinfo, JPEG_APP0 + 2, 0xFFFF);
2797 prior to calling jpeg_read_header(). jpeg_read_icc_profile() can be called at
2798 any point between jpeg_read_header() and jpeg_finish_decompress().
2801 Raw (downsampled) image data
2802 ----------------------------
2804 Some applications need to supply already-downsampled image data to the JPEG
2805 compressor, or to receive raw downsampled data from the decompressor. The
2806 library supports this requirement by allowing the application to write or
2807 read raw data, bypassing the normal preprocessing or postprocessing steps.
2808 The interface is different from the standard one and is somewhat harder to
2809 use. If your interest is merely in bypassing color conversion, we recommend
2810 that you use the standard interface and simply set jpeg_color_space =
2811 in_color_space (or jpeg_color_space = out_color_space for decompression).
2812 The mechanism described in this section is necessary only to supply or
2813 receive downsampled image data, in which not all components have the same
2817 To compress raw data, you must supply the data in the colorspace to be used
2818 in the JPEG file (please read the earlier section on Special color spaces)
2819 and downsampled to the sampling factors specified in the JPEG parameters.
2820 You must supply the data in the format used internally by the JPEG library,
2821 namely a J*SAMPIMAGE array. This is an array of pointers to two-dimensional
2822 arrays, each of type J*SAMPARRAY. Each 2-D array holds the values for one
2823 color component. This structure is necessary since the components are of
2824 different sizes. If the image dimensions are not a multiple of the MCU size,
2825 you must also pad the data correctly (usually, this is done by replicating
2826 the last column and/or row). The data must be padded to a multiple of a DCT
2827 block in each component: that is, each downsampled row must contain a
2828 multiple of 8 valid samples, and there must be a multiple of 8 sample rows
2829 for each component. (For applications such as conversion of digital TV
2830 images, the standard image size is usually a multiple of the DCT block size,
2831 so that no padding need actually be done.)
2833 The procedure for compression of raw data is basically the same as normal
2834 compression, except that you call jpeg_write_raw_data() or
2835 jpeg12_write_raw_data() in place of jpeg_write_scanlines() or
2836 jpeg12_write_scanlines(). Before calling jpeg_start_compress(), you must do
2838 * Set cinfo->raw_data_in to TRUE. (It is set FALSE by jpeg_set_defaults().)
2839 This notifies the library that you will be supplying raw data.
2840 * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
2841 call is a good idea. Note that since color conversion is bypassed,
2842 in_color_space is ignored, except that jpeg_set_defaults() uses it to
2843 choose the default jpeg_color_space setting.
2844 * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
2845 cinfo->comp_info[i].v_samp_factor, are correct. Since these indicate the
2846 dimensions of the data you are supplying, it's wise to set them
2847 explicitly, rather than assuming the library's defaults are what you want.
2849 To pass raw data to the library, call jpeg*_write_raw_data() in place of
2850 jpeg*_write_scanlines(). The routines work similarly except that
2851 jpeg*_write_raw_data takes a J*SAMPIMAGE data array rather than J*SAMPARRAY.
2852 The scanlines count passed to and returned from jpeg*_write_raw_data is
2853 measured in terms of the component with the largest v_samp_factor.
2855 jpeg*_write_raw_data() processes one MCU row per call, which is to say
2856 v_samp_factor*DCTSIZE sample rows of each component. The passed num_lines
2857 value must be at least max_v_samp_factor*DCTSIZE, and the return value will
2858 be exactly that amount (or possibly some multiple of that amount, in future
2859 library versions). This is true even on the last call at the bottom of the
2860 image; don't forget to pad your data as necessary.
2862 The required dimensions of the supplied data can be computed for each
2864 cinfo->comp_info[i].width_in_blocks*DCTSIZE samples per row
2865 cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
2866 after jpeg_start_compress() has initialized those fields. If the valid data
2867 is smaller than this, it must be padded appropriately. For some sampling
2868 factors and image sizes, additional dummy DCT blocks are inserted to make
2869 the image a multiple of the MCU dimensions. The library creates such dummy
2870 blocks itself; it does not read them from your supplied data. Therefore you
2871 need never pad by more than DCTSIZE samples. An example may help here.
2872 Assume 2h2v downsampling of YCbCr data, that is
2873 cinfo->comp_info[0].h_samp_factor = 2 for Y
2874 cinfo->comp_info[0].v_samp_factor = 2
2875 cinfo->comp_info[1].h_samp_factor = 1 for Cb
2876 cinfo->comp_info[1].v_samp_factor = 1
2877 cinfo->comp_info[2].h_samp_factor = 1 for Cr
2878 cinfo->comp_info[2].v_samp_factor = 1
2879 and suppose that the nominal image dimensions (cinfo->image_width and
2880 cinfo->image_height) are 101x101 pixels. Then jpeg_start_compress() will
2881 compute downsampled_width = 101 and width_in_blocks = 13 for Y,
2882 downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
2883 for the height fields). You must pad the Y data to at least 13*8 = 104
2884 columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows. The
2885 MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
2886 scanlines on each call to jpeg*_write_raw_data(), which is to say 16 actual
2887 sample rows of Y and 8 each of Cb and Cr. A total of 7 MCU rows are needed,
2888 so you must pass a total of 7*16 = 112 "scanlines". The last DCT block row
2889 of Y data is dummy, so it doesn't matter what you pass for it in the data
2890 arrays, but the scanlines count must total up to 112 so that all of the Cb
2891 and Cr data gets passed.
2893 Output suspension is supported with raw-data compression: if the data
2894 destination module suspends, jpeg*_write_raw_data() will return 0.
2895 In this case the same data rows must be passed again on the next call.
2898 Decompression with raw data output implies bypassing all postprocessing:
2899 you cannot ask for rescaling or color quantization, for instance. More
2900 seriously, you must deal with the color space and sampling factors present in
2901 the incoming file. If your application only handles, say, 2h1v YCbCr data,
2902 you must check for and fail on other color spaces or other sampling factors.
2903 The library will not convert to a different color space for you.
2905 To obtain raw data output, set cinfo->raw_data_out = TRUE before
2906 jpeg_start_decompress() (it is set FALSE by jpeg_read_header()). Be sure to
2907 verify that the color space and sampling factors are ones you can handle.
2908 Then call jpeg_read_raw_data() or jpeg12_read_raw_data() in place of
2909 jpeg_read_scanlines() or jpeg12_read_scanlines(). The decompression process is
2910 otherwise the same as usual.
2912 jpeg*_read_raw_data() returns one MCU row per call, and thus you must pass a
2913 buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
2914 the same as for raw-data compression). The buffer you pass must be large
2915 enough to hold the actual data plus padding to DCT-block boundaries. As with
2916 compression, any entirely dummy DCT blocks are not processed so you need not
2917 allocate space for them, but the total scanline count includes them. The
2918 above example of computing buffer dimensions for raw-data compression is
2919 equally valid for decompression.
2921 Input suspension is supported with raw-data decompression: if the data source
2922 module suspends, jpeg*_read_raw_data() will return 0. You can also use
2923 buffered-image mode to read raw data in multiple passes.
2926 Really raw data: DCT coefficients
2927 ---------------------------------
2929 It is possible to read or write the contents of a JPEG file as raw DCT
2930 coefficients. This facility is mainly intended for use in lossless
2931 transcoding between different JPEG file formats. Other possible applications
2932 include lossless cropping of a JPEG image, lossless reassembly of a
2933 multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
2935 To read the contents of a JPEG file as DCT coefficients, open the file and do
2936 jpeg_read_header() as usual. But instead of calling jpeg_start_decompress()
2937 and jpeg*_read_scanlines(), call jpeg_read_coefficients(). This will read the
2938 entire image into a set of virtual coefficient-block arrays, one array per
2939 component. The return value is a pointer to an array of virtual-array
2940 descriptors. Each virtual array can be accessed directly using the JPEG
2941 memory manager's access_virt_barray method (see Memory management, below,
2942 and also read structure.txt's discussion of virtual array handling). Or,
2943 for simple transcoding to a different JPEG file format, the array list can
2944 just be handed directly to jpeg_write_coefficients().
2946 Each block in the block arrays contains quantized coefficient values in
2947 normal array order (not JPEG zigzag order). The block arrays contain only
2948 DCT blocks containing real data; any entirely-dummy blocks added to fill out
2949 interleaved MCUs at the right or bottom edges of the image are discarded
2950 during reading and are not stored in the block arrays. (The size of each
2951 block array can be determined from the width_in_blocks and height_in_blocks
2952 fields of the component's comp_info entry.) This is also the data format
2953 expected by jpeg_write_coefficients().
2955 When you are done using the virtual arrays, call jpeg_finish_decompress()
2956 to release the array storage and return the decompression object to an idle
2957 state; or just call jpeg_destroy() if you don't need to reuse the object.
2959 If you use a suspending data source, jpeg_read_coefficients() will return
2960 NULL if it is forced to suspend; a non-NULL return value indicates successful
2961 completion. You need not test for a NULL return value when using a
2962 non-suspending data source.
2964 It is also possible to call jpeg_read_coefficients() to obtain access to the
2965 decoder's coefficient arrays during a normal decode cycle in buffered-image
2966 mode. This frammish might be useful for progressively displaying an incoming
2967 image and then re-encoding it without loss. To do this, decode in buffered-
2968 image mode as discussed previously, then call jpeg_read_coefficients() after
2969 the last jpeg_finish_output() call. The arrays will be available for your use
2970 until you call jpeg_finish_decompress().
2973 To write the contents of a JPEG file as DCT coefficients, you must provide
2974 the DCT coefficients stored in virtual block arrays. You can either pass
2975 block arrays read from an input JPEG file by jpeg_read_coefficients(), or
2976 allocate virtual arrays from the JPEG compression object and fill them
2977 yourself. In either case, jpeg_write_coefficients() is substituted for
2978 jpeg_start_compress() and jpeg*_write_scanlines(). Thus the sequence is
2979 * Create compression object
2980 * Set all compression parameters as necessary
2981 * Request virtual arrays if needed
2982 * jpeg_write_coefficients()
2983 * jpeg_finish_compress()
2984 * Destroy or re-use compression object
2985 jpeg_write_coefficients() is passed a pointer to an array of virtual block
2986 array descriptors; the number of arrays is equal to cinfo.num_components.
2988 The virtual arrays need only have been requested, not realized, before
2989 jpeg_write_coefficients() is called. A side-effect of
2990 jpeg_write_coefficients() is to realize any virtual arrays that have been
2991 requested from the compression object's memory manager. Thus, when obtaining
2992 the virtual arrays from the compression object, you should fill the arrays
2993 after calling jpeg_write_coefficients(). The data is actually written out
2994 when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
2997 When writing raw DCT coefficients, it is crucial that the JPEG quantization
2998 tables and sampling factors match the way the data was encoded, or the
2999 resulting file will be invalid. For transcoding from an existing JPEG file,
3000 we recommend using jpeg_copy_critical_parameters(). This routine initializes
3001 all the compression parameters to default values (like jpeg_set_defaults()),
3002 then copies the critical information from a source decompression object.
3003 The decompression object should have just been used to read the entire
3004 JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
3006 jpeg_write_coefficients() marks all tables stored in the compression object
3007 as needing to be written to the output file (thus, it acts like
3008 jpeg_start_compress(cinfo, TRUE)). This is for safety's sake, to avoid
3009 emitting abbreviated JPEG files by accident. If you really want to emit an
3010 abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
3011 individual sent_table flags, between calling jpeg_write_coefficients() and
3012 jpeg_finish_compress().
3018 Some applications may need to regain control from the JPEG library every so
3019 often. The typical use of this feature is to produce a percent-done bar or
3020 other progress display. (For a simple example, see cjpeg.c or djpeg.c.)
3021 Although you do get control back frequently during the data-transferring pass
3022 (the jpeg*_read_scanlines or jpeg*_write_scanlines loop), any additional passes
3023 will occur inside jpeg_finish_compress or jpeg_start_decompress; those
3024 routines may take a long time to execute, and you don't get control back
3025 until they are done.
3027 You can define a progress-monitor routine which will be called periodically
3028 by the library. No guarantees are made about how often this call will occur,
3029 so we don't recommend you use it for mouse tracking or anything like that.
3030 At present, a call will occur once per MCU row, scanline, or sample row
3031 group, whichever unit is convenient for the current processing mode; so the
3032 wider the image, the longer the time between calls. During the data
3033 transferring pass, only one call occurs per call of jpeg*_read_scanlines or
3034 jpeg*_write_scanlines, so don't pass a large number of scanlines at once if
3035 you want fine resolution in the progress count. (If you really need to use
3036 the callback mechanism for time-critical tasks like mouse tracking, you could
3037 insert additional calls inside some of the library's inner loops.)
3039 To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
3040 fill in its progress_monitor field with a pointer to your callback routine,
3041 and set cinfo->progress to point to the struct. The callback will be called
3042 whenever cinfo->progress is non-NULL. (This pointer is set to NULL by
3043 jpeg_create_compress or jpeg_create_decompress; the library will not change
3044 it thereafter. So if you allocate dynamic storage for the progress struct,
3045 make sure it will live as long as the JPEG object does. Allocating from the
3046 JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.) You
3047 can use the same callback routine for both compression and decompression.
3049 The jpeg_progress_mgr struct contains four fields which are set by the library:
3050 long pass_counter; /* work units completed in this pass */
3051 long pass_limit; /* total number of work units in this pass */
3052 int completed_passes; /* passes completed so far */
3053 int total_passes; /* total number of passes expected */
3054 During any one pass, pass_counter increases from 0 up to (not including)
3055 pass_limit; the step size is usually but not necessarily 1. The pass_limit
3056 value may change from one pass to another. The expected total number of
3057 passes is in total_passes, and the number of passes already completed is in
3058 completed_passes. Thus the fraction of work completed may be estimated as
3059 completed_passes + (pass_counter/pass_limit)
3060 --------------------------------------------
3062 ignoring the fact that the passes may not be equal amounts of work.
3064 When decompressing, pass_limit can even change within a pass, because it
3065 depends on the number of scans in the JPEG file, which isn't always known in
3066 advance. The computed fraction-of-work-done may jump suddenly (if the library
3067 discovers it has overestimated the number of scans) or even decrease (in the
3068 opposite case). It is not wise to put great faith in the work estimate.
3070 When using the decompressor's buffered-image mode, the progress monitor work
3071 estimate is likely to be completely unhelpful, because the library has no way
3072 to know how many output passes will be demanded of it. Currently, the library
3073 sets total_passes based on the assumption that there will be one more output
3074 pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
3075 TRUE), but no more output passes if the file end has been reached when the
3076 output pass is started. This means that total_passes will rise as additional
3077 output passes are requested. If you have a way of determining the input file
3078 size, estimating progress based on the fraction of the file that's been read
3079 will probably be more useful than using the library's value.
3085 This section covers some key facts about the JPEG library's built-in memory
3086 manager. For more info, please read structure.txt's section about the memory
3087 manager, and consult the source code if necessary.
3089 All memory and temporary file allocation within the library is done via the
3090 memory manager. If necessary, you can replace the "back end" of the memory
3091 manager to control allocation yourself (for example, if you don't want the
3092 library to use malloc() and free() for some reason).
3094 Some data is allocated "permanently" and will not be freed until the JPEG
3095 object is destroyed. Most data is allocated "per image" and is freed by
3096 jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort. You can call the
3097 memory manager yourself to allocate structures that will automatically be
3098 freed at these times. Typical code for this is
3099 ptr = (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE, size);
3100 Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
3101 Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
3102 There are also alloc_sarray and alloc_barray routines that automatically
3103 build 2-D sample or block arrays.
3105 The library's minimum space requirements to process an image depend on the
3106 image's width, but not on its height, because the library ordinarily works
3107 with "strip" buffers that are as wide as the image but just a few rows high.
3108 Some operating modes (eg, two-pass color quantization) require full-image
3109 buffers. Such buffers are treated as "virtual arrays": only the current strip
3110 need be in memory, and the rest can be swapped out to a temporary file.
3112 When using temporary files, the library will make the in-memory buffers for
3113 its virtual arrays just big enough to stay within a "maximum memory" setting.
3114 Your application can set this limit by setting cinfo->mem->max_memory_to_use
3115 after creating the JPEG object. (Of course, there is still a minimum size for
3116 the buffers, so the max-memory setting is effective only if it is bigger than
3117 the minimum space needed.) If you allocate any large structures yourself, you
3118 must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
3119 order to have them counted against the max memory limit. Also keep in mind
3120 that space allocated with alloc_small() is ignored, on the assumption that
3121 it's too small to be worth worrying about; so a reasonable safety margin
3122 should be left when setting max_memory_to_use.
3124 NOTE: Unless you develop your own memory manager back end, then temporary files
3125 will never be used. The back end provided in libjpeg-turbo (jmemnobs.c) simply
3126 malloc()s and free()s virtual arrays, and an error occurs if the required
3127 memory exceeds the limit specified in cinfo->mem->max_memory_to_use.
3133 Working memory requirements while performing compression or decompression
3134 depend on image dimensions, image characteristics (such as colorspace and
3135 JPEG process), and operating mode (application-selected options).
3137 As of v6b, the decompressor requires:
3138 1. About 24K in more-or-less-fixed-size data. This varies a bit depending
3139 on operating mode and image characteristics (particularly color vs.
3140 grayscale), but it doesn't depend on image dimensions.
3141 2. Strip buffers (of size proportional to the image width) for IDCT and
3142 upsampling results. The worst case for commonly used sampling factors
3143 is about 34 bytes * width in pixels for a color image. A grayscale image
3144 only needs about 8 bytes per pixel column.
3145 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
3146 file (including progressive JPEGs), or whenever you select buffered-image
3147 mode. This takes 2 bytes/coefficient. At typical 2x2 sampling, that's
3148 3 bytes per pixel for a color image. Worst case (1x1 sampling) requires
3149 6 bytes/pixel. For grayscale, figure 2 bytes/pixel.
3150 4. To perform 2-pass color quantization, the decompressor also needs a
3151 128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
3152 This does not count any memory allocated by the application, such as a
3153 buffer to hold the final output image.
3155 The above figures are valid for 8-bit JPEG data precision and a machine with
3156 32-bit ints. For 12-bit and 16-bit JPEG data, double the size of the strip
3157 buffers and quantization pixel buffer. The "fixed-size" data will be somewhat
3158 smaller with 16-bit ints, larger with 64-bit ints. Also, CMYK or other unusual
3159 color spaces will require different amounts of space.
3161 The full-image coefficient and pixel buffers, if needed at all, do not
3162 have to be fully RAM resident; you can have the library use temporary
3163 files instead when the total memory usage would exceed a limit you set.
3164 (But if your OS supports virtual memory, it's probably better to just use
3165 jmemnobs and let the OS do the swapping.)
3167 The compressor's memory requirements are similar, except that it has no need
3168 for color quantization. Also, it needs a full-image DCT coefficient buffer
3169 if Huffman-table optimization is asked for, even if progressive mode is not
3172 If you need more detailed information about memory usage in a particular
3173 situation, you can enable the MEM_STATS code in jmemmgr.c.
3176 Library compile-time options
3177 ----------------------------
3179 A number of compile-time options are available by modifying jmorecfg.h.
3181 The maximum number of components (color channels) in the image is determined
3182 by MAX_COMPONENTS. The JPEG standard allows up to 255 components, but we
3183 expect that few applications will need more than four or so.
3185 On machines with unusual data type sizes, you may be able to improve
3186 performance or reduce memory space by tweaking the various typedefs in
3187 jmorecfg.h. In particular, on some RISC CPUs, access to arrays of "short"s
3188 is quite slow; consider trading memory for speed by making JCOEF, INT16, and
3189 UINT16 be "int" or "unsigned int". UINT8 is also a candidate to become int.
3190 You probably don't want to make J*SAMPLE be int unless you have lots of memory
3193 You can reduce the size of the library by compiling out various optional
3194 functions. To do this, undefine xxx_SUPPORTED symbols as necessary.
3196 You can also save a few K by not having text error messages in the library;
3197 the standard error message table occupies about 5Kb. This is particularly
3198 reasonable for embedded applications where there's no good way to display
3199 a message anyway. To do this, remove the creation of the message table
3200 (jpeg_std_message_table[]) from jerror.c, and alter format_message to do
3201 something reasonable without it. You could output the numeric value of the
3202 message code number, for example. If you do this, you can also save a couple
3203 more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
3204 you don't need trace capability anyway, right?
3207 Portability considerations
3208 --------------------------
3210 The JPEG library has been written to be extremely portable; the sample
3211 applications cjpeg and djpeg are slightly less so. This section summarizes
3212 the design goals in this area. (If you encounter any bugs that cause the
3213 library to be less portable than is claimed here, we'd appreciate hearing
3216 The code works fine on ANSI C and C++ compilers, using any of the popular
3217 system include file setups, and some not-so-popular ones too.
3219 The code is not dependent on the exact sizes of the C data types. As
3220 distributed, we make the assumptions that
3221 char is at least 8 bits wide
3222 short is at least 16 bits wide
3223 int is at least 16 bits wide
3224 long is at least 32 bits wide
3225 (These are the minimum requirements of the ANSI C standard.) Wider types will
3226 work fine, although memory may be used inefficiently if char is much larger
3227 than 8 bits or short is much bigger than 16 bits. The code should work
3228 equally well with 16- or 32-bit ints.
3230 In a system where these assumptions are not met, you may be able to make the
3231 code work by modifying the typedefs in jmorecfg.h. However, you will probably
3232 have difficulty if int is less than 16 bits wide, since references to plain
3233 int abound in the code.
3235 char can be either signed or unsigned, although the code runs faster if an
3236 unsigned char type is available. If char is wider than 8 bits, you will need
3237 to redefine JOCTET and/or provide custom data source/destination managers so
3238 that JOCTET represents exactly 8 bits of data on external storage.
3240 The JPEG library proper does not assume ASCII representation of characters.
3241 But some of the image file I/O modules in cjpeg/djpeg do have ASCII
3242 dependencies in file-header manipulation; so does cjpeg's select_file_type()
3245 The JPEG library does not rely heavily on the C library. In particular, C
3246 stdio is used only by the data source/destination modules and the error
3247 handler, all of which are application-replaceable. (cjpeg/djpeg are more
3248 heavily dependent on stdio.) malloc and free are called only from the memory
3249 manager "back end" module, so you can use a different memory allocator by
3250 replacing that one file.
3252 More info about porting the code may be gleaned by reading jconfig.txt,
3253 jmorecfg.h, and jinclude.h.