2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
12 /* This is a simple program that reads ivf files and decodes them
13 * using the new interface. Decoded frames are output as YV12 raw.
22 #define snprintf _snprintf
23 #define isatty _isatty
24 #define fileno _fileno
28 #define VPX_CODEC_DISABLE_COMPAT 1
29 #include "vpx_config.h"
30 #include "vpx/vpx_decoder.h"
31 #include "vpx_ports/vpx_timer.h"
32 #if CONFIG_VP8_DECODER
33 #include "vpx/vp8dx.h"
36 #include "md5_utils.h"
38 #include "nestegg/include/nestegg/nestegg.h"
44 static const char *exec_name;
46 #define VP8_FOURCC (0x00385056)
50 const vpx_codec_iface_t *iface;
52 unsigned int fourcc_mask;
55 #if CONFIG_VP8_DECODER
56 {"vp8", &vpx_codec_vp8_dx_algo, VP8_FOURCC, 0x00FFFFFF},
61 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
63 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
64 "Output raw YV12 frames");
65 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
66 "Output raw I420 frames");
67 static const arg_def_t flipuvarg = ARG_DEF(NULL, "flipuv", 0,
68 "Flip the chroma planes in the output");
69 static const arg_def_t noblitarg = ARG_DEF(NULL, "noblit", 0,
70 "Don't process the decoded frames");
71 static const arg_def_t progressarg = ARG_DEF(NULL, "progress", 0,
72 "Show progress after each frame decodes");
73 static const arg_def_t limitarg = ARG_DEF(NULL, "limit", 1,
74 "Stop decoding after n frames");
75 static const arg_def_t postprocarg = ARG_DEF(NULL, "postproc", 0,
76 "Postprocess decoded frames");
77 static const arg_def_t summaryarg = ARG_DEF(NULL, "summary", 0,
78 "Show timing summary");
79 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
80 "Output file name pattern (see below)");
81 static const arg_def_t threadsarg = ARG_DEF("t", "threads", 1,
82 "Max threads to use");
83 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
84 "Show version string");
87 static const arg_def_t md5arg = ARG_DEF(NULL, "md5", 0,
88 "Compute the MD5 sum of the decoded frame");
90 static const arg_def_t *all_args[] =
92 &codecarg, &use_yv12, &use_i420, &flipuvarg, &noblitarg,
93 &progressarg, &limitarg, &postprocarg, &summaryarg, &outputfile,
94 &threadsarg, &verbosearg,
101 #if CONFIG_VP8_DECODER
102 static const arg_def_t addnoise_level = ARG_DEF(NULL, "noise-level", 1,
103 "Enable VP8 postproc add noise");
104 static const arg_def_t deblock = ARG_DEF(NULL, "deblock", 0,
105 "Enable VP8 deblocking");
106 static const arg_def_t demacroblock_level = ARG_DEF(NULL, "demacroblock-level", 1,
107 "Enable VP8 demacroblocking, w/ level");
108 static const arg_def_t pp_debug_info = ARG_DEF(NULL, "pp-debug-info", 1,
109 "Enable VP8 visible debug info");
112 static const arg_def_t *vp8_pp_args[] =
114 &addnoise_level, &deblock, &demacroblock_level, &pp_debug_info,
119 static void usage_exit()
123 fprintf(stderr, "Usage: %s <options> filename\n\n"
124 "Options:\n", exec_name);
125 arg_show_usage(stderr, all_args);
126 #if CONFIG_VP8_DECODER
127 fprintf(stderr, "\nVP8 Postprocessing Options:\n");
128 arg_show_usage(stderr, vp8_pp_args);
131 "\nOutput File Patterns:\n\n"
132 " The -o argument specifies the name of the file(s) to "
133 "write to. If the\n argument does not include any escape "
134 "characters, the output will be\n written to a single file. "
135 "Otherwise, the filename will be calculated by\n expanding "
136 "the following escape characters:\n"
137 "\n\t%%w - Frame width"
138 "\n\t%%h - Frame height"
139 "\n\t%%<n> - Frame number, zero padded to <n> places (1..9)"
140 "\n\n Pattern arguments are only supported in conjunction "
141 "with the --yv12 and\n --i420 options. If the -o option is "
142 "not specified, the output will be\n directed to stdout.\n"
144 fprintf(stderr, "\nIncluded decoders:\n\n");
146 for (i = 0; i < sizeof(ifaces) / sizeof(ifaces[0]); i++)
147 fprintf(stderr, " %-6s - %s\n",
149 vpx_codec_iface_name(ifaces[i].iface));
154 void die(const char *fmt, ...)
158 vfprintf(stderr, fmt, ap);
159 fprintf(stderr, "\n");
163 static unsigned int mem_get_le16(const void *vmem)
166 const unsigned char *mem = (const unsigned char *)vmem;
173 static unsigned int mem_get_le32(const void *vmem)
176 const unsigned char *mem = (const unsigned char *)vmem;
196 nestegg *nestegg_ctx;
200 unsigned int video_track;
203 #define IVF_FRAME_HDR_SZ (sizeof(uint32_t) + sizeof(uint64_t))
204 #define RAW_FRAME_HDR_SZ (sizeof(uint32_t))
205 static int read_frame(struct input_ctx *input,
208 size_t *buf_alloc_sz)
210 char raw_hdr[IVF_FRAME_HDR_SZ];
212 FILE *infile = input->infile;
213 enum file_kind kind = input->kind;
214 if(kind == WEBM_FILE)
216 if(input->chunk >= input->chunks)
222 /* End of this packet, get another. */
224 nestegg_free_packet(input->pkt);
226 if(nestegg_read_packet(input->nestegg_ctx, &input->pkt) <= 0
227 || nestegg_packet_track(input->pkt, &track))
230 } while(track != input->video_track);
232 if(nestegg_packet_count(input->pkt, &input->chunks))
237 if(nestegg_packet_data(input->pkt, input->chunk, buf, buf_sz))
243 /* For both the raw and ivf formats, the frame size is the first 4 bytes
244 * of the frame header. We just need to special case on the header
247 else if (fread(raw_hdr, kind==IVF_FILE
248 ? IVF_FRAME_HDR_SZ : RAW_FRAME_HDR_SZ, 1, infile) != 1)
251 fprintf(stderr, "Failed to read frame size\n");
257 new_buf_sz = mem_get_le32(raw_hdr);
259 if (new_buf_sz > 256 * 1024 * 1024)
261 fprintf(stderr, "Error: Read invalid frame size (%u)\n",
262 (unsigned int)new_buf_sz);
266 if (kind == RAW_FILE && new_buf_sz > 256 * 1024)
267 fprintf(stderr, "Warning: Read invalid frame size (%u)"
268 " - not a raw file?\n", (unsigned int)new_buf_sz);
270 if (new_buf_sz > *buf_alloc_sz)
272 uint8_t *new_buf = realloc(*buf, 2 * new_buf_sz);
277 *buf_alloc_sz = 2 * new_buf_sz;
281 fprintf(stderr, "Failed to allocate compressed data buffer\n");
287 *buf_sz = new_buf_sz;
291 if (fread(*buf, 1, *buf_sz, infile) != *buf_sz)
293 fprintf(stderr, "Failed to read full frame\n");
303 void *out_open(const char *out_fn, int do_md5)
310 MD5Context *md5_ctx = out = malloc(sizeof(MD5Context));
317 FILE *outfile = out = strcmp("-", out_fn) ? fopen(out_fn, "wb") : stdout;
321 fprintf(stderr, "Failed to output file");
329 void out_put(void *out, const uint8_t *buf, unsigned int len, int do_md5)
334 MD5Update(out, buf, len);
339 if(fwrite(buf, 1, len, out));
343 void out_close(void *out, const char *out_fn, int do_md5)
354 for (i = 0; i < 16; i++)
355 printf("%02x", md5[i]);
357 printf(" %s\n", out_fn);
366 unsigned int file_is_ivf(FILE *infile,
367 unsigned int *fourcc,
369 unsigned int *height,
370 unsigned int *fps_den,
371 unsigned int *fps_num)
376 if (fread(raw_hdr, 1, 32, infile) == 32)
378 if (raw_hdr[0] == 'D' && raw_hdr[1] == 'K'
379 && raw_hdr[2] == 'I' && raw_hdr[3] == 'F')
383 if (mem_get_le16(raw_hdr + 4) != 0)
384 fprintf(stderr, "Error: Unrecognized IVF version! This file may not"
385 " decode properly.");
387 *fourcc = mem_get_le32(raw_hdr + 8);
388 *width = mem_get_le16(raw_hdr + 12);
389 *height = mem_get_le16(raw_hdr + 14);
390 *fps_num = mem_get_le32(raw_hdr + 16);
391 *fps_den = mem_get_le32(raw_hdr + 20);
393 /* Some versions of vpxenc used 1/(2*fps) for the timebase, so
394 * we can guess the framerate using only the timebase in this
395 * case. Other files would require reading ahead to guess the
396 * timebase, like we do for webm.
400 /* Correct for the factor of 2 applied to the timebase in the
403 if(*fps_num&1)*fps_den<<=1;
408 /* Don't know FPS for sure, and don't have readahead code
409 * (yet?), so just default to 30fps.
424 unsigned int file_is_raw(FILE *infile,
425 unsigned int *fourcc,
427 unsigned int *height,
428 unsigned int *fps_den,
429 unsigned int *fps_num)
431 unsigned char buf[32];
433 vpx_codec_stream_info_t si;
435 if (fread(buf, 1, 32, infile) == 32)
439 if(mem_get_le32(buf) < 256 * 1024 * 1024)
440 for (i = 0; i < sizeof(ifaces) / sizeof(ifaces[0]); i++)
441 if(!vpx_codec_peek_stream_info(ifaces[i].iface,
442 buf + 4, 32 - 4, &si))
445 *fourcc = ifaces[i].fourcc;
460 nestegg_read_cb(void *buffer, size_t length, void *userdata)
464 if(fread(buffer, 1, length, f) < length)
476 nestegg_seek_cb(int64_t offset, int whence, void * userdata)
479 case NESTEGG_SEEK_SET: whence = SEEK_SET; break;
480 case NESTEGG_SEEK_CUR: whence = SEEK_CUR; break;
481 case NESTEGG_SEEK_END: whence = SEEK_END; break;
483 return fseek(userdata, offset, whence)? -1 : 0;
488 nestegg_tell_cb(void * userdata)
490 return ftell(userdata);
495 nestegg_log_cb(nestegg * context, unsigned int severity, char const * format,
500 va_start(ap, format);
501 vfprintf(stderr, format, ap);
502 fprintf(stderr, "\n");
508 webm_guess_framerate(struct input_ctx *input,
509 unsigned int *fps_den,
510 unsigned int *fps_num)
515 /* Guess the framerate. Read up to 1 second, or 50 video packets,
516 * whichever comes first.
518 for(i=0; tstamp < 1000000000 && i < 50;)
520 nestegg_packet * pkt;
523 if(nestegg_read_packet(input->nestegg_ctx, &pkt) <= 0)
526 nestegg_packet_track(pkt, &track);
527 if(track == input->video_track)
529 nestegg_packet_tstamp(pkt, &tstamp);
533 nestegg_free_packet(pkt);
536 if(nestegg_track_seek(input->nestegg_ctx, input->video_track, 0))
539 *fps_num = (i - 1) * 1000000;
540 *fps_den = tstamp / 1000;
543 input->nestegg_ctx = NULL;
544 rewind(input->infile);
550 file_is_webm(struct input_ctx *input,
551 unsigned int *fourcc,
553 unsigned int *height,
554 unsigned int *fps_den,
555 unsigned int *fps_num)
561 nestegg_io io = {nestegg_read_cb, nestegg_seek_cb, nestegg_tell_cb,
563 nestegg_video_params params;
564 nestegg_packet * pkt;
566 if(nestegg_init(&input->nestegg_ctx, io, NULL))
569 if(nestegg_track_count(input->nestegg_ctx, &n))
574 track_type = nestegg_track_type(input->nestegg_ctx, i);
576 if(track_type == NESTEGG_TRACK_VIDEO)
578 else if(track_type < 0)
582 if(nestegg_track_codec_id(input->nestegg_ctx, i) != NESTEGG_CODEC_VP8)
584 fprintf(stderr, "Not VP8 video, quitting.\n");
588 input->video_track = i;
590 if(nestegg_track_video_params(input->nestegg_ctx, i, ¶ms))
595 *fourcc = VP8_FOURCC;
596 *width = params.width;
597 *height = params.height;
600 input->nestegg_ctx = NULL;
601 rewind(input->infile);
606 void show_progress(int frame_in, int frame_out, unsigned long dx_time)
608 fprintf(stderr, "%d decoded frames/%d showed frames in %lu us (%.2f fps)\r",
609 frame_in, frame_out, dx_time,
610 (float)frame_out * 1000000.0 / (float)dx_time);
614 void generate_filename(const char *pattern, char *out, size_t q_len,
615 unsigned int d_w, unsigned int d_h,
616 unsigned int frame_in)
618 const char *p = pattern;
623 char *next_pat = strchr(p, '%');
633 case 'w': snprintf(q, q_len - 1, "%d", d_w); break;
634 case 'h': snprintf(q, q_len - 1, "%d", d_h); break;
635 case '1': snprintf(q, q_len - 1, "%d", frame_in); break;
636 case '2': snprintf(q, q_len - 1, "%02d", frame_in); break;
637 case '3': snprintf(q, q_len - 1, "%03d", frame_in); break;
638 case '4': snprintf(q, q_len - 1, "%04d", frame_in); break;
639 case '5': snprintf(q, q_len - 1, "%05d", frame_in); break;
640 case '6': snprintf(q, q_len - 1, "%06d", frame_in); break;
641 case '7': snprintf(q, q_len - 1, "%07d", frame_in); break;
642 case '8': snprintf(q, q_len - 1, "%08d", frame_in); break;
643 case '9': snprintf(q, q_len - 1, "%09d", frame_in); break;
645 die("Unrecognized pattern %%%c\n", p[1]);
649 if(pat_len >= q_len - 1)
650 die("Output filename too long.\n");
659 // copy the next segment
661 copy_len = strlen(p);
663 copy_len = next_pat - p;
665 if(copy_len >= q_len - 1)
666 die("Output filename too long.\n");
668 memcpy(q, p, copy_len);
678 int main(int argc, const char **argv_)
680 vpx_codec_ctx_t decoder;
684 size_t buf_sz = 0, buf_alloc_sz = 0;
686 int frame_in = 0, frame_out = 0, flipuv = 0, noblit = 0, do_md5 = 0, progress = 0;
687 int stop_after = 0, postproc = 0, summary = 0, quiet = 1;
688 vpx_codec_iface_t *iface = NULL;
690 unsigned long dx_time = 0;
692 char **argv, **argi, **argj;
693 const char *outfile_pattern = 0;
694 char outfile[PATH_MAX];
699 unsigned int fps_den;
700 unsigned int fps_num;
702 vpx_codec_dec_cfg_t cfg = {0};
703 #if CONFIG_VP8_DECODER
704 vp8_postproc_cfg_t vp8_pp_cfg = {0};
706 struct input_ctx input = {0};
708 /* Parse command line */
709 exec_name = argv_[0];
710 argv = argv_dup(argc - 1, argv_ + 1);
712 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
714 memset(&arg, 0, sizeof(arg));
717 if (arg_match(&arg, &codecarg, argi))
721 for (j = 0; j < sizeof(ifaces) / sizeof(ifaces[0]); j++)
722 if (!strcmp(ifaces[j].name, arg.val))
726 iface = ifaces[k].iface;
728 die("Error: Unrecognized argument (%s) to --codec\n",
731 else if (arg_match(&arg, &outputfile, argi))
732 outfile_pattern = arg.val;
733 else if (arg_match(&arg, &use_yv12, argi))
738 else if (arg_match(&arg, &use_i420, argi))
743 else if (arg_match(&arg, &flipuvarg, argi))
745 else if (arg_match(&arg, &noblitarg, argi))
747 else if (arg_match(&arg, &progressarg, argi))
749 else if (arg_match(&arg, &limitarg, argi))
750 stop_after = arg_parse_uint(&arg);
751 else if (arg_match(&arg, &postprocarg, argi))
753 else if (arg_match(&arg, &md5arg, argi))
755 else if (arg_match(&arg, &summaryarg, argi))
757 else if (arg_match(&arg, &threadsarg, argi))
758 cfg.threads = arg_parse_uint(&arg);
759 else if (arg_match(&arg, &verbosearg, argi))
762 #if CONFIG_VP8_DECODER
763 else if (arg_match(&arg, &addnoise_level, argi))
766 vp8_pp_cfg.post_proc_flag |= VP8_ADDNOISE;
767 vp8_pp_cfg.noise_level = arg_parse_uint(&arg);
769 else if (arg_match(&arg, &demacroblock_level, argi))
772 vp8_pp_cfg.post_proc_flag |= VP8_DEMACROBLOCK;
773 vp8_pp_cfg.deblocking_level = arg_parse_uint(&arg);
775 else if (arg_match(&arg, &deblock, argi))
778 vp8_pp_cfg.post_proc_flag |= VP8_DEBLOCK;
780 else if (arg_match(&arg, &pp_debug_info, argi))
782 unsigned int level = arg_parse_uint(&arg);
785 vp8_pp_cfg.post_proc_flag &= ~0x7;
788 vp8_pp_cfg.post_proc_flag |= level;
796 /* Check for unrecognized options */
797 for (argi = argv; *argi; argi++)
798 if (argi[0][0] == '-' && strlen(argi[0]) > 1)
799 die("Error: Unrecognized option %s\n", *argi);
801 /* Handle non-option arguments */
808 infile = strcmp(fn, "-") ? fopen(fn, "rb") : stdin;
812 fprintf(stderr, "Failed to open file '%s'",
813 strcmp(fn, "-") ? fn : "stdin");
817 /* Make sure we don't dump to the terminal, unless forced to with -o - */
818 if(!outfile_pattern && isatty(fileno(stdout)) && !do_md5 && !noblit)
821 "Not dumping raw video to your terminal. Use '-o -' to "
826 input.infile = infile;
827 if(file_is_ivf(infile, &fourcc, &width, &height, &fps_den,
829 input.kind = IVF_FILE;
830 else if(file_is_webm(&input, &fourcc, &width, &height, &fps_den, &fps_num))
831 input.kind = WEBM_FILE;
832 else if(file_is_raw(infile, &fourcc, &width, &height, &fps_den, &fps_num))
833 input.kind = RAW_FILE;
836 fprintf(stderr, "Unrecognized input file type.\n");
840 /* If the output file is not set or doesn't have a sequence number in
841 * it, then we only open it once.
843 outfile_pattern = outfile_pattern ? outfile_pattern : "-";
846 const char *p = outfile_pattern;
850 if(p && p[1] >= '1' && p[1] <= '9')
852 // pattern contains sequence number, so it's not unique.
861 if(single_file && !noblit)
863 generate_filename(outfile_pattern, outfile, sizeof(outfile)-1,
865 out = out_open(outfile, do_md5);
868 if (use_y4m && !noblit)
873 fprintf(stderr, "YUV4MPEG2 not supported with output patterns,"
874 " try --i420 or --yv12.\n");
878 if(input.kind == WEBM_FILE)
879 webm_guess_framerate(&input, &fps_den, &fps_num);
881 /*Note: We can't output an aspect ratio here because IVF doesn't
882 store one, and neither does VP8.
883 That will have to wait until these tools support WebM natively.*/
884 sprintf(buffer, "YUV4MPEG2 C%s W%u H%u F%u:%u I%c\n",
885 "420jpeg", width, height, fps_num, fps_den, 'p');
886 out_put(out, (unsigned char *)buffer, strlen(buffer), do_md5);
889 /* Try to determine the codec from the fourcc. */
890 for (i = 0; i < sizeof(ifaces) / sizeof(ifaces[0]); i++)
891 if ((fourcc & ifaces[i].fourcc_mask) == ifaces[i].fourcc)
893 vpx_codec_iface_t *ivf_iface = ifaces[i].iface;
895 if (iface && iface != ivf_iface)
896 fprintf(stderr, "Notice -- IVF header indicates codec: %s\n",
904 if (vpx_codec_dec_init(&decoder, iface ? iface : ifaces[0].iface, &cfg,
905 postproc ? VPX_CODEC_USE_POSTPROC : 0))
907 fprintf(stderr, "Failed to initialize decoder: %s\n", vpx_codec_error(&decoder));
912 fprintf(stderr, "%s\n", decoder.name);
914 #if CONFIG_VP8_DECODER
916 if (vp8_pp_cfg.post_proc_flag
917 && vpx_codec_control(&decoder, VP8_SET_POSTPROC, &vp8_pp_cfg))
919 fprintf(stderr, "Failed to configure postproc: %s\n", vpx_codec_error(&decoder));
926 while (!read_frame(&input, &buf, &buf_sz, &buf_alloc_sz))
928 vpx_codec_iter_t iter = NULL;
930 struct vpx_usec_timer timer;
932 vpx_usec_timer_start(&timer);
934 if (vpx_codec_decode(&decoder, buf, buf_sz, NULL, 0))
936 const char *detail = vpx_codec_error_detail(&decoder);
937 fprintf(stderr, "Failed to decode frame: %s\n", vpx_codec_error(&decoder));
940 fprintf(stderr, " Additional information: %s\n", detail);
945 vpx_usec_timer_mark(&timer);
946 dx_time += vpx_usec_timer_elapsed(&timer);
950 if ((img = vpx_codec_get_frame(&decoder, &iter)))
954 show_progress(frame_in, frame_out, dx_time);
961 char out_fn[PATH_MAX];
966 size_t len = sizeof(out_fn)-1;
969 generate_filename(outfile_pattern, out_fn, len-1,
970 img->d_w, img->d_h, frame_in);
971 out = out_open(out_fn, do_md5);
974 out_put(out, (unsigned char *)"FRAME\n", 6, do_md5);
976 buf = img->planes[VPX_PLANE_Y];
978 for (y = 0; y < img->d_h; y++)
980 out_put(out, buf, img->d_w, do_md5);
981 buf += img->stride[VPX_PLANE_Y];
984 buf = img->planes[flipuv?VPX_PLANE_V:VPX_PLANE_U];
986 for (y = 0; y < (1 + img->d_h) / 2; y++)
988 out_put(out, buf, (1 + img->d_w) / 2, do_md5);
989 buf += img->stride[VPX_PLANE_U];
992 buf = img->planes[flipuv?VPX_PLANE_U:VPX_PLANE_V];
994 for (y = 0; y < (1 + img->d_h) / 2; y++)
996 out_put(out, buf, (1 + img->d_w) / 2, do_md5);
997 buf += img->stride[VPX_PLANE_V];
1001 out_close(out, out_fn, do_md5);
1005 if (stop_after && frame_in >= stop_after)
1009 if (summary || progress)
1011 show_progress(frame_in, frame_out, dx_time);
1012 fprintf(stderr, "\n");
1017 if (vpx_codec_destroy(&decoder))
1019 fprintf(stderr, "Failed to destroy decoder: %s\n", vpx_codec_error(&decoder));
1020 return EXIT_FAILURE;
1023 if (single_file && !noblit)
1024 out_close(out, outfile, do_md5);
1026 if(input.nestegg_ctx)
1027 nestegg_destroy(input.nestegg_ctx);
1028 if(input.kind != WEBM_FILE)
1033 return EXIT_SUCCESS;