/* * Copyright (c) 2010 The WebM project authors. All Rights Reserved. * * Use of this source code is governed by a BSD-style license * that can be found in the LICENSE file in the root of the source * tree. An additional intellectual property rights grant can be found * in the file PATENTS. All contributing project authors may * be found in the AUTHORS file in the root of the source tree. */ /* This is a simple program that reads ivf files and decodes them * using the new interface. Decoded frames are output as YV12 raw. */ #include #include #include #include #define VPX_CODEC_DISABLE_COMPAT 1 #include "vpx_config.h" #include "vpx/vpx_decoder.h" #include "vpx_ports/vpx_timer.h" #if CONFIG_VP8_DECODER #include "vpx/vp8dx.h" #endif #if CONFIG_MD5 #include "md5_utils.h" #endif #include "nestegg/include/nestegg/nestegg.h" static const char *exec_name; #define VP8_FOURCC (0x00385056) static const struct { char const *name; const vpx_codec_iface_t *iface; unsigned int fourcc; unsigned int fourcc_mask; } ifaces[] = { #if CONFIG_VP8_DECODER {"vp8", &vpx_codec_vp8_dx_algo, VP8_FOURCC, 0x00FFFFFF}, #endif }; #include "args.h" static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use"); static const arg_def_t prefixarg = ARG_DEF("p", "prefix", 1, "Prefix to use when saving frames"); static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0, "Output file is YV12 "); static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0, "Output file is I420 (default)"); static const arg_def_t flipuvarg = ARG_DEF(NULL, "flipuv", 0, "Synonym for --yv12"); static const arg_def_t noblitarg = ARG_DEF(NULL, "noblit", 0, "Don't process the decoded frames"); static const arg_def_t progressarg = ARG_DEF(NULL, "progress", 0, "Show progress after each frame decodes"); static const arg_def_t limitarg = ARG_DEF(NULL, "limit", 1, "Stop decoding after n frames"); static const arg_def_t postprocarg = ARG_DEF(NULL, "postproc", 0, "Postprocess decoded frames"); static const arg_def_t summaryarg = ARG_DEF(NULL, "summary", 0, "Show timing summary"); static const arg_def_t outputfile = ARG_DEF("o", "output", 1, "Output raw yv12 file instead of images"); static const arg_def_t usey4marg = ARG_DEF("y", "y4m", 0, "Output file is YUV4MPEG2"); static const arg_def_t threadsarg = ARG_DEF("t", "threads", 1, "Max threads to use"); static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0, "Suppress version string"); #if CONFIG_MD5 static const arg_def_t md5arg = ARG_DEF(NULL, "md5", 0, "Compute the MD5 sum of the decoded frame"); #endif static const arg_def_t *all_args[] = { &codecarg, &prefixarg, &use_yv12, &use_i420, &flipuvarg, &noblitarg, &progressarg, &limitarg, &postprocarg, &summaryarg, &outputfile, &usey4marg, &threadsarg, &quietarg, #if CONFIG_MD5 &md5arg, #endif NULL }; #if CONFIG_VP8_DECODER static const arg_def_t addnoise_level = ARG_DEF(NULL, "noise-level", 1, "Enable VP8 postproc add noise"); static const arg_def_t deblock = ARG_DEF(NULL, "deblock", 0, "Enable VP8 deblocking"); static const arg_def_t demacroblock_level = ARG_DEF(NULL, "demacroblock-level", 1, "Enable VP8 demacroblocking, w/ level"); static const arg_def_t pp_debug_info = ARG_DEF(NULL, "pp-debug-info", 1, "Enable VP8 visible debug info"); static const arg_def_t *vp8_pp_args[] = { &addnoise_level, &deblock, &demacroblock_level, &pp_debug_info, NULL }; #endif static void usage_exit() { int i; fprintf(stderr, "Usage: %s filename\n\n" "Options:\n", exec_name); arg_show_usage(stderr, all_args); #if CONFIG_VP8_DECODER fprintf(stderr, "\nVP8 Postprocessing Options:\n"); arg_show_usage(stderr, vp8_pp_args); #endif fprintf(stderr, "\nIncluded decoders:\n\n"); for (i = 0; i < sizeof(ifaces) / sizeof(ifaces[0]); i++) fprintf(stderr, " %-6s - %s\n", ifaces[i].name, vpx_codec_iface_name(ifaces[i].iface)); exit(EXIT_FAILURE); } void die(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vfprintf(stderr, fmt, ap); fprintf(stderr, "\n"); usage_exit(); } static unsigned int mem_get_le16(const void *vmem) { unsigned int val; const unsigned char *mem = (const unsigned char *)vmem; val = mem[1] << 8; val |= mem[0]; return val; } static unsigned int mem_get_le32(const void *vmem) { unsigned int val; const unsigned char *mem = (const unsigned char *)vmem; val = mem[3] << 24; val |= mem[2] << 16; val |= mem[1] << 8; val |= mem[0]; return val; } enum file_kind { RAW_FILE, IVF_FILE, WEBM_FILE }; struct input_ctx { enum file_kind kind; FILE *infile; nestegg *nestegg_ctx; nestegg_packet *pkt; unsigned int chunk; unsigned int chunks; unsigned int video_track; }; #define IVF_FRAME_HDR_SZ (sizeof(uint32_t) + sizeof(uint64_t)) #define RAW_FRAME_HDR_SZ (sizeof(uint32_t)) static int read_frame(struct input_ctx *input, uint8_t **buf, size_t *buf_sz, size_t *buf_alloc_sz) { char raw_hdr[IVF_FRAME_HDR_SZ]; size_t new_buf_sz; FILE *infile = input->infile; enum file_kind kind = input->kind; if(kind == WEBM_FILE) { if(input->chunk >= input->chunks) { unsigned int track; do { /* End of this packet, get another. */ if(input->pkt) nestegg_free_packet(input->pkt); if(nestegg_read_packet(input->nestegg_ctx, &input->pkt) <= 0 || nestegg_packet_track(input->pkt, &track)) return 1; } while(track != input->video_track); if(nestegg_packet_count(input->pkt, &input->chunks)) return 1; input->chunk = 0; } if(nestegg_packet_data(input->pkt, input->chunk, buf, buf_sz)) return 1; input->chunk++; return 0; } /* For both the raw and ivf formats, the frame size is the first 4 bytes * of the frame header. We just need to special case on the header * size. */ else if (fread(raw_hdr, kind==IVF_FILE ? IVF_FRAME_HDR_SZ : RAW_FRAME_HDR_SZ, 1, infile) != 1) { if (!feof(infile)) fprintf(stderr, "Failed to read frame size\n"); new_buf_sz = 0; } else { new_buf_sz = mem_get_le32(raw_hdr); if (new_buf_sz > 256 * 1024 * 1024) { fprintf(stderr, "Error: Read invalid frame size (%u)\n", (unsigned int)new_buf_sz); new_buf_sz = 0; } if (kind == RAW_FILE && new_buf_sz > 256 * 1024) fprintf(stderr, "Warning: Read invalid frame size (%u)" " - not a raw file?\n", (unsigned int)new_buf_sz); if (new_buf_sz > *buf_alloc_sz) { uint8_t *new_buf = realloc(*buf, 2 * new_buf_sz); if (new_buf) { *buf = new_buf; *buf_alloc_sz = 2 * new_buf_sz; } else { fprintf(stderr, "Failed to allocate compressed data buffer\n"); new_buf_sz = 0; } } } *buf_sz = new_buf_sz; if (*buf_sz) { if (fread(*buf, 1, *buf_sz, infile) != *buf_sz) { fprintf(stderr, "Failed to read full frame\n"); return 1; } return 0; } return 1; } void *out_open(const char *out_fn, int do_md5) { void *out = NULL; if (do_md5) { #if CONFIG_MD5 MD5Context *md5_ctx = out = malloc(sizeof(MD5Context)); (void)out_fn; MD5Init(md5_ctx); #endif } else { FILE *outfile = out = strcmp("-", out_fn) ? fopen(out_fn, "wb") : stdout; if (!outfile) { fprintf(stderr, "Failed to output file"); exit(EXIT_FAILURE); } } return out; } void out_put(void *out, const uint8_t *buf, unsigned int len, int do_md5) { if (do_md5) { #if CONFIG_MD5 MD5Update(out, buf, len); #endif } else { fwrite(buf, 1, len, out); } } void out_close(void *out, const char *out_fn, int do_md5) { if (do_md5) { #if CONFIG_MD5 uint8_t md5[16]; int i; MD5Final(md5, out); free(out); for (i = 0; i < 16; i++) printf("%02x", md5[i]); printf(" %s\n", out_fn); #endif } else { fclose(out); } } unsigned int file_is_ivf(FILE *infile, unsigned int *fourcc, unsigned int *width, unsigned int *height, unsigned int *fps_den, unsigned int *fps_num) { char raw_hdr[32]; int is_ivf = 0; if (fread(raw_hdr, 1, 32, infile) == 32) { if (raw_hdr[0] == 'D' && raw_hdr[1] == 'K' && raw_hdr[2] == 'I' && raw_hdr[3] == 'F') { is_ivf = 1; if (mem_get_le16(raw_hdr + 4) != 0) fprintf(stderr, "Error: Unrecognized IVF version! This file may not" " decode properly."); *fourcc = mem_get_le32(raw_hdr + 8); *width = mem_get_le16(raw_hdr + 12); *height = mem_get_le16(raw_hdr + 14); *fps_num = mem_get_le32(raw_hdr + 16); *fps_den = mem_get_le32(raw_hdr + 20); /* Some versions of ivfenc used 1/(2*fps) for the timebase, so * we can guess the framerate using only the timebase in this * case. Other files would require reading ahead to guess the * timebase, like we do for webm. */ if(*fps_num < 1000) { /* Correct for the factor of 2 applied to the timebase in the * encoder. */ if(*fps_num&1)*fps_den<<=1; else *fps_num>>=1; } else { /* Don't know FPS for sure, and don't have readahead code * (yet?), so just default to 30fps. */ *fps_num = 30; *fps_den = 1; } } } if (!is_ivf) rewind(infile); return is_ivf; } static int nestegg_read_cb(void *buffer, size_t length, void *userdata) { FILE *f = userdata; fread(buffer, 1, length, f); if (ferror(f)) return -1; if (feof(f)) return 0; return 1; } static int nestegg_seek_cb(int64_t offset, int whence, void * userdata) { switch(whence) { case NESTEGG_SEEK_SET: whence = SEEK_SET; break; case NESTEGG_SEEK_CUR: whence = SEEK_CUR; break; case NESTEGG_SEEK_END: whence = SEEK_END; break; }; return fseek(userdata, offset, whence)? -1 : 0; } static int64_t nestegg_tell_cb(void * userdata) { return ftell(userdata); } static void nestegg_log_cb(nestegg * context, unsigned int severity, char const * format, ...) { va_list ap; va_start(ap, format); vfprintf(stderr, format, ap); fprintf(stderr, "\n"); va_end(ap); } static int webm_guess_framerate(struct input_ctx *input, unsigned int *fps_den, unsigned int *fps_num) { unsigned int i; uint64_t tstamp=0; /* Guess the framerate. Read up to 1 second, or 50 video packets, * whichever comes first. */ for(i=0; tstamp < 1000000000 && i < 50;) { nestegg_packet * pkt; unsigned int track; if(nestegg_read_packet(input->nestegg_ctx, &pkt) <= 0) break; nestegg_packet_track(pkt, &track); if(track == input->video_track) { nestegg_packet_tstamp(pkt, &tstamp); i++; } nestegg_free_packet(pkt); } if(nestegg_track_seek(input->nestegg_ctx, input->video_track, 0)) goto fail; *fps_num = (i - 1) * 1000000; *fps_den = tstamp / 1000; return 0; fail: input->nestegg_ctx = NULL; rewind(input->infile); return 1; } static int file_is_webm(struct input_ctx *input, unsigned int *fourcc, unsigned int *width, unsigned int *height, unsigned int *fps_den, unsigned int *fps_num) { unsigned int i, n; int track_type = -1; uint64_t tstamp=0; nestegg_io io = {nestegg_read_cb, nestegg_seek_cb, nestegg_tell_cb, input->infile}; nestegg_video_params params; nestegg_packet * pkt; if(nestegg_init(&input->nestegg_ctx, io, NULL)) goto fail; if(nestegg_track_count(input->nestegg_ctx, &n)) goto fail; for(i=0; inestegg_ctx, i); if(track_type == NESTEGG_TRACK_VIDEO) break; else if(track_type < 0) goto fail; } if(nestegg_track_codec_id(input->nestegg_ctx, i) != NESTEGG_CODEC_VP8) { fprintf(stderr, "Not VP8 video, quitting.\n"); exit(1); } input->video_track = i; if(nestegg_track_video_params(input->nestegg_ctx, i, ¶ms)) goto fail; *fps_den = 0; *fps_num = 0; *fourcc = VP8_FOURCC; *width = params.width; *height = params.height; return 1; fail: input->nestegg_ctx = NULL; rewind(input->infile); return 0; } int main(int argc, const char **argv_) { vpx_codec_ctx_t decoder; char *prefix = NULL, *fn = NULL; int i; uint8_t *buf = NULL; size_t buf_sz = 0, buf_alloc_sz = 0; FILE *infile; int frame_in = 0, frame_out = 0, flipuv = 0, noblit = 0, do_md5 = 0, progress = 0; int stop_after = 0, postproc = 0, summary = 0, quiet = 0; vpx_codec_iface_t *iface = NULL; unsigned int fourcc; unsigned long dx_time = 0; struct arg arg; char **argv, **argi, **argj; const char *fn2 = 0; int use_y4m = 0; unsigned int width; unsigned int height; unsigned int fps_den; unsigned int fps_num; void *out = NULL; vpx_codec_dec_cfg_t cfg = {0}; #if CONFIG_VP8_DECODER vp8_postproc_cfg_t vp8_pp_cfg = {0}; #endif struct input_ctx input = {0}; /* Parse command line */ exec_name = argv_[0]; argv = argv_dup(argc - 1, argv_ + 1); for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) { memset(&arg, 0, sizeof(arg)); arg.argv_step = 1; if (arg_match(&arg, &codecarg, argi)) { int j, k = -1; for (j = 0; j < sizeof(ifaces) / sizeof(ifaces[0]); j++) if (!strcmp(ifaces[j].name, arg.val)) k = j; if (k >= 0) iface = ifaces[k].iface; else die("Error: Unrecognized argument (%s) to --codec\n", arg.val); } else if (arg_match(&arg, &outputfile, argi)) fn2 = arg.val; else if (arg_match(&arg, &usey4marg, argi)) use_y4m = 1; else if (arg_match(&arg, &prefixarg, argi)) prefix = strdup(arg.val); else if (arg_match(&arg, &use_yv12, argi)) flipuv = 1; else if (arg_match(&arg, &use_i420, argi)) flipuv = 0; else if (arg_match(&arg, &flipuvarg, argi)) flipuv = 1; else if (arg_match(&arg, &noblitarg, argi)) noblit = 1; else if (arg_match(&arg, &progressarg, argi)) progress = 1; else if (arg_match(&arg, &limitarg, argi)) stop_after = arg_parse_uint(&arg); else if (arg_match(&arg, &postprocarg, argi)) postproc = 1; else if (arg_match(&arg, &md5arg, argi)) do_md5 = 1; else if (arg_match(&arg, &summaryarg, argi)) summary = 1; else if (arg_match(&arg, &threadsarg, argi)) cfg.threads = arg_parse_uint(&arg); else if (arg_match(&arg, &quietarg, argi)) quiet = 1; #if CONFIG_VP8_DECODER else if (arg_match(&arg, &addnoise_level, argi)) { postproc = 1; vp8_pp_cfg.post_proc_flag |= VP8_ADDNOISE; vp8_pp_cfg.noise_level = arg_parse_uint(&arg); } else if (arg_match(&arg, &demacroblock_level, argi)) { postproc = 1; vp8_pp_cfg.post_proc_flag |= VP8_DEMACROBLOCK; vp8_pp_cfg.deblocking_level = arg_parse_uint(&arg); } else if (arg_match(&arg, &deblock, argi)) { postproc = 1; vp8_pp_cfg.post_proc_flag |= VP8_DEBLOCK; } else if (arg_match(&arg, &pp_debug_info, argi)) { unsigned int level = arg_parse_uint(&arg); postproc = 1; vp8_pp_cfg.post_proc_flag &= ~0x7; if (level) vp8_pp_cfg.post_proc_flag |= 8 << (level - 1); } #endif else argj++; } /* Check for unrecognized options */ for (argi = argv; *argi; argi++) if (argi[0][0] == '-' && strlen(argi[0]) > 1) die("Error: Unrecognized option %s\n", *argi); /* Handle non-option arguments */ fn = argv[0]; if (!fn) usage_exit(); if (!prefix) prefix = strdup("img"); /* Open file */ infile = strcmp(fn, "-") ? fopen(fn, "rb") : stdin; if (!infile) { fprintf(stderr, "Failed to open file"); return EXIT_FAILURE; } if (fn2) out = out_open(fn2, do_md5); input.infile = infile; if(file_is_ivf(infile, &fourcc, &width, &height, &fps_den, &fps_num)) input.kind = IVF_FILE; else if(file_is_webm(&input, &fourcc, &width, &height, &fps_den, &fps_num)) input.kind = WEBM_FILE; else input.kind = RAW_FILE; if (input.kind != RAW_FILE) { if (use_y4m) { char buffer[128]; if (!fn2) { fprintf(stderr, "YUV4MPEG2 output only supported with -o.\n"); return EXIT_FAILURE; } if(input.kind == WEBM_FILE) webm_guess_framerate(&input, &fps_den, &fps_num); /*Note: We can't output an aspect ratio here because IVF doesn't store one, and neither does VP8. That will have to wait until these tools support WebM natively.*/ sprintf(buffer, "YUV4MPEG2 C%s W%u H%u F%u:%u I%c\n", "420jpeg", width, height, fps_num, fps_den, 'p'); out_put(out, (unsigned char *)buffer, strlen(buffer), do_md5); } /* Try to determine the codec from the fourcc. */ for (i = 0; i < sizeof(ifaces) / sizeof(ifaces[0]); i++) if ((fourcc & ifaces[i].fourcc_mask) == ifaces[i].fourcc) { vpx_codec_iface_t *ivf_iface = ifaces[i].iface; if (iface && iface != ivf_iface) fprintf(stderr, "Notice -- IVF header indicates codec: %s\n", ifaces[i].name); else iface = ivf_iface; break; } } else if(use_y4m) { fprintf(stderr, "YUV4MPEG2 output not supported from raw input.\n"); return EXIT_FAILURE; } if (vpx_codec_dec_init(&decoder, iface ? iface : ifaces[0].iface, &cfg, postproc ? VPX_CODEC_USE_POSTPROC : 0)) { fprintf(stderr, "Failed to initialize decoder: %s\n", vpx_codec_error(&decoder)); return EXIT_FAILURE; } if (!quiet) fprintf(stderr, "%s\n", decoder.name); #if CONFIG_VP8_DECODER if (vp8_pp_cfg.post_proc_flag && vpx_codec_control(&decoder, VP8_SET_POSTPROC, &vp8_pp_cfg)) { fprintf(stderr, "Failed to configure postproc: %s\n", vpx_codec_error(&decoder)); return EXIT_FAILURE; } #endif /* Decode file */ while (!read_frame(&input, &buf, &buf_sz, &buf_alloc_sz)) { vpx_codec_iter_t iter = NULL; vpx_image_t *img; struct vpx_usec_timer timer; vpx_usec_timer_start(&timer); if (vpx_codec_decode(&decoder, buf, buf_sz, NULL, 0)) { const char *detail = vpx_codec_error_detail(&decoder); fprintf(stderr, "Failed to decode frame: %s\n", vpx_codec_error(&decoder)); if (detail) fprintf(stderr, " Additional information: %s\n", detail); goto fail; } vpx_usec_timer_mark(&timer); dx_time += vpx_usec_timer_elapsed(&timer); ++frame_in; if (progress) fprintf(stderr, "decoded frame %d.\n", frame_in); if ((img = vpx_codec_get_frame(&decoder, &iter))) ++frame_out; if (!noblit) { if (img) { unsigned int y; char out_fn[128+24]; uint8_t *buf; const char *sfx = flipuv ? "yv12" : "i420"; if (!fn2) { sprintf(out_fn, "%s-%dx%d-%04d.%s", prefix, img->d_w, img->d_h, frame_in, sfx); out = out_open(out_fn, do_md5); } else if(use_y4m) out_put(out, (unsigned char *)"FRAME\n", 6, do_md5); buf = img->planes[VPX_PLANE_Y]; for (y = 0; y < img->d_h; y++) { out_put(out, buf, img->d_w, do_md5); buf += img->stride[VPX_PLANE_Y]; } buf = img->planes[flipuv?VPX_PLANE_V:VPX_PLANE_U]; for (y = 0; y < (1 + img->d_h) / 2; y++) { out_put(out, buf, (1 + img->d_w) / 2, do_md5); buf += img->stride[VPX_PLANE_U]; } buf = img->planes[flipuv?VPX_PLANE_U:VPX_PLANE_V]; for (y = 0; y < (1 + img->d_h) / 2; y++) { out_put(out, buf, (1 + img->d_w) / 2, do_md5); buf += img->stride[VPX_PLANE_V]; } if (!fn2) out_close(out, out_fn, do_md5); } } if (stop_after && frame_in >= stop_after) break; } if (summary) { fprintf(stderr, "%d decoded frames/%d showed frames in %lu us (%.2f fps)\n", frame_in, frame_out, dx_time, (float)frame_out * 1000000.0 / (float)dx_time); } fail: if (vpx_codec_destroy(&decoder)) { fprintf(stderr, "Failed to destroy decoder: %s\n", vpx_codec_error(&decoder)); return EXIT_FAILURE; } if (fn2) out_close(out, fn2, do_md5); if(input.nestegg_ctx) nestegg_destroy(input.nestegg_ctx); if(input.kind != WEBM_FILE) free(buf); fclose(infile); free(prefix); free(argv); return EXIT_SUCCESS; }