static int encode_frame(vpx_codec_ctx_t *codec, vpx_image_t *img, int frame_index, VpxVideoWriter *writer) { int got_pkts = 0; vpx_codec_iter_t iter = NULL; const vpx_codec_cx_pkt_t *pkt = NULL; const vpx_codec_err_t res = vpx_codec_encode(codec, img, frame_index, 1, 0, VPX_DL_GOOD_QUALITY); if (res != VPX_CODEC_OK) die_codec(codec, "Failed to encode frame"); while ((pkt = vpx_codec_get_cx_data(codec, &iter)) != NULL) { got_pkts = 1; if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) { const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0; if (!vpx_video_writer_write_frame(writer, pkt->data.frame.buf, pkt->data.frame.sz, pkt->data.frame.pts)) { die_codec(codec, "Failed to write compressed frame"); } printf(keyframe ? "K" : "."); fflush(stdout); } } return got_pkts; }
static int encode_frame(vpx_codec_ctx_t *ctx, const vpx_image_t *img, vpx_codec_pts_t pts, unsigned int duration, vpx_enc_frame_flags_t flags, unsigned int deadline, VpxVideoWriter *writer) { int got_pkts = 0; vpx_codec_iter_t iter = NULL; const vpx_codec_cx_pkt_t *pkt = NULL; const vpx_codec_err_t res = vpx_codec_encode(ctx, img, pts, duration, flags, deadline); if (res != VPX_CODEC_OK) die_codec(ctx, "Failed to encode frame."); while ((pkt = vpx_codec_get_cx_data(ctx, &iter)) != NULL) { got_pkts = 1; if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) { const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0; if (!vpx_video_writer_write_frame(writer, pkt->data.frame.buf, pkt->data.frame.sz, pkt->data.frame.pts)) die_codec(ctx, "Failed to write compressed frame."); printf(keyframe ? "K" : "."); fflush(stdout); } } return got_pkts; }
static int encode_frame(vpx_codec_ctx_t *ecodec, vpx_codec_enc_cfg_t *cfg, vpx_image_t *img, unsigned int frame_in, VpxVideoWriter *writer, int test_decode, vpx_codec_ctx_t *dcodec, unsigned int *frame_out, int *mismatch_seen) { int got_pkts = 0; vpx_codec_iter_t iter = NULL; const vpx_codec_cx_pkt_t *pkt = NULL; int got_data; const vpx_codec_err_t res = vpx_codec_encode(ecodec, img, frame_in, 1, 0, VPX_DL_GOOD_QUALITY); if (res != VPX_CODEC_OK) die_codec(ecodec, "Failed to encode frame"); got_data = 0; while ((pkt = vpx_codec_get_cx_data(ecodec, &iter)) != NULL) { got_pkts = 1; if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) { const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0; if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) { *frame_out += 1; } if (!vpx_video_writer_write_frame(writer, pkt->data.frame.buf, pkt->data.frame.sz, pkt->data.frame.pts)) { die_codec(ecodec, "Failed to write compressed frame"); } printf(keyframe ? "K" : "."); fflush(stdout); got_data = 1; // Decode 1 frame. if (test_decode) { if (vpx_codec_decode(dcodec, pkt->data.frame.buf, (unsigned int)pkt->data.frame.sz, NULL, 0)) die_codec(dcodec, "Failed to decode frame."); } } } // Mismatch checking if (got_data && test_decode) { testing_decode(ecodec, dcodec, cfg, *frame_out, mismatch_seen); } return got_pkts; }
int main(int argc, char **argv) { VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL}; vpx_codec_ctx_t codec; vpx_codec_enc_cfg_t cfg; int frame_cnt = 0; vpx_image_t raw; vpx_codec_err_t res; unsigned int width; unsigned int height; int speed; int frame_avail; int got_data; int flags = 0; unsigned int i; int pts = 0; // PTS starts at 0. int frame_duration = 1; // 1 timebase tick per frame. int layering_mode = 0; int layer_flags[VPX_TS_MAX_PERIODICITY] = {0}; int flag_periodicity = 1; vpx_svc_layer_id_t layer_id = {0, 0}; const VpxInterface *encoder = NULL; FILE *infile = NULL; struct RateControlMetrics rc; int64_t cx_time = 0; const int min_args_base = 11; #if CONFIG_VP9_HIGHBITDEPTH vpx_bit_depth_t bit_depth = VPX_BITS_8; int input_bit_depth = 8; const int min_args = min_args_base + 1; #else const int min_args = min_args_base; #endif // CONFIG_VP9_HIGHBITDEPTH double sum_bitrate = 0.0; double sum_bitrate2 = 0.0; double framerate = 30.0; exec_name = argv[0]; // Check usage and arguments. if (argc < min_args) { #if CONFIG_VP9_HIGHBITDEPTH die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> " "<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> " "<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n", argv[0]); #else die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> " "<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> " "<Rate_0> ... <Rate_nlayers-1> \n", argv[0]); #endif // CONFIG_VP9_HIGHBITDEPTH } encoder = get_vpx_encoder_by_name(argv[3]); if (!encoder) die("Unsupported codec."); printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface())); width = strtol(argv[4], NULL, 0); height = strtol(argv[5], NULL, 0); if (width < 16 || width % 2 || height < 16 || height % 2) { die("Invalid resolution: %d x %d", width, height); } layering_mode = strtol(argv[10], NULL, 0); if (layering_mode < 0 || layering_mode > 12) { die("Invalid layering mode (0..12) %s", argv[10]); } if (argc != min_args + mode_to_num_layers[layering_mode]) { die("Invalid number of arguments"); } #if CONFIG_VP9_HIGHBITDEPTH switch (strtol(argv[argc-1], NULL, 0)) { case 8: bit_depth = VPX_BITS_8; input_bit_depth = 8; break; case 10: bit_depth = VPX_BITS_10; input_bit_depth = 10; break; case 12: bit_depth = VPX_BITS_12; input_bit_depth = 12; break; default: die("Invalid bit depth (8, 10, 12) %s", argv[argc-1]); } if (!vpx_img_alloc(&raw, bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016, width, height, 32)) { die("Failed to allocate image", width, height); } #else if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) { die("Failed to allocate image", width, height); } #endif // CONFIG_VP9_HIGHBITDEPTH // Populate encoder configuration. res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0); if (res) { printf("Failed to get config: %s\n", vpx_codec_err_to_string(res)); return EXIT_FAILURE; } // Update the default configuration with our settings. cfg.g_w = width; cfg.g_h = height; #if CONFIG_VP9_HIGHBITDEPTH if (bit_depth != VPX_BITS_8) { cfg.g_bit_depth = bit_depth; cfg.g_input_bit_depth = input_bit_depth; cfg.g_profile = 2; } #endif // CONFIG_VP9_HIGHBITDEPTH // Timebase format e.g. 30fps: numerator=1, demoninator = 30. cfg.g_timebase.num = strtol(argv[6], NULL, 0); cfg.g_timebase.den = strtol(argv[7], NULL, 0); speed = strtol(argv[8], NULL, 0); if (speed < 0) { die("Invalid speed setting: must be positive"); } for (i = min_args_base; (int)i < min_args_base + mode_to_num_layers[layering_mode]; ++i) { cfg.ts_target_bitrate[i - 11] = strtol(argv[i], NULL, 0); } // Real time parameters. cfg.rc_dropframe_thresh = strtol(argv[9], NULL, 0); cfg.rc_end_usage = VPX_CBR; cfg.rc_resize_allowed = 0; cfg.rc_min_quantizer = 2; cfg.rc_max_quantizer = 56; cfg.rc_undershoot_pct = 50; cfg.rc_overshoot_pct = 50; cfg.rc_buf_initial_sz = 500; cfg.rc_buf_optimal_sz = 600; cfg.rc_buf_sz = 1000; // Enable error resilient mode. cfg.g_error_resilient = 1; cfg.g_lag_in_frames = 0; cfg.kf_mode = VPX_KF_AUTO; // Disable automatic keyframe placement. cfg.kf_min_dist = cfg.kf_max_dist = 3000; set_temporal_layer_pattern(layering_mode, &cfg, layer_flags, &flag_periodicity); set_rate_control_metrics(&rc, &cfg); // Target bandwidth for the whole stream. // Set to ts_target_bitrate for highest layer (total bitrate). cfg.rc_target_bitrate = cfg.ts_target_bitrate[cfg.ts_number_layers - 1]; // Open input file. if (!(infile = fopen(argv[1], "rb"))) { die("Failed to open %s for reading", argv[1]); } framerate = cfg.g_timebase.den / cfg.g_timebase.num; // Open an output file for each stream. for (i = 0; i < cfg.ts_number_layers; ++i) { char file_name[PATH_MAX]; VpxVideoInfo info; info.codec_fourcc = encoder->fourcc; info.frame_width = cfg.g_w; info.frame_height = cfg.g_h; info.time_base.numerator = cfg.g_timebase.num; info.time_base.denominator = cfg.g_timebase.den; snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i); outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info); if (!outfile[i]) die("Failed to open %s for writing", file_name); assert(outfile[i] != NULL); } // No spatial layers in this encoder. cfg.ss_number_layers = 1; // Initialize codec. #if CONFIG_VP9_HIGHBITDEPTH if (vpx_codec_enc_init( &codec, encoder->codec_interface(), &cfg, bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH)) #else if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0)) #endif // CONFIG_VP9_HIGHBITDEPTH die_codec(&codec, "Failed to initialize encoder"); if (strncmp(encoder->name, "vp8", 3) == 0) { vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed); vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kDenoiserOnYOnly); } else if (strncmp(encoder->name, "vp9", 3) == 0) { vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed); vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3); vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0); vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, 0); if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1: 0)) { die_codec(&codec, "Failed to set SVC"); } } vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1); vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1); // This controls the maximum target size of the key frame. // For generating smaller key frames, use a smaller max_intra_size_pct // value, like 100 or 200. { const int max_intra_size_pct = 900; vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT, max_intra_size_pct); } frame_avail = 1; while (frame_avail || got_data) { struct vpx_usec_timer timer; vpx_codec_iter_t iter = NULL; const vpx_codec_cx_pkt_t *pkt; // Update the temporal layer_id. No spatial layers in this test. layer_id.spatial_layer_id = 0; layer_id.temporal_layer_id = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity]; if (strncmp(encoder->name, "vp9", 3) == 0) { vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id); } flags = layer_flags[frame_cnt % flag_periodicity]; frame_avail = vpx_img_read(&raw, infile); if (frame_avail) ++rc.layer_input_frames[layer_id.temporal_layer_id]; vpx_usec_timer_start(&timer); if (vpx_codec_encode(&codec, frame_avail? &raw : NULL, pts, 1, flags, VPX_DL_REALTIME)) { die_codec(&codec, "Failed to encode frame"); } vpx_usec_timer_mark(&timer); cx_time += vpx_usec_timer_elapsed(&timer); // Reset KF flag. if (layering_mode != 7) { layer_flags[0] &= ~VPX_EFLAG_FORCE_KF; } got_data = 0; while ( (pkt = vpx_codec_get_cx_data(&codec, &iter)) ) { got_data = 1; switch (pkt->kind) { case VPX_CODEC_CX_FRAME_PKT: for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity]; i < cfg.ts_number_layers; ++i) { vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf, pkt->data.frame.sz, pts); ++rc.layer_tot_enc_frames[i]; rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz; // Keep count of rate control stats per layer (for non-key frames). if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] && !(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) { rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz; rc.layer_avg_rate_mismatch[i] += fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) / rc.layer_pfb[i]; ++rc.layer_enc_frames[i]; } } // Update for short-time encoding bitrate states, for moving window // of size rc->window, shifted by rc->window / 2. // Ignore first window segment, due to key frame. if (frame_cnt > rc.window_size) { sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate; if (frame_cnt % rc.window_size == 0) { rc.window_count += 1; rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size; rc.variance_st_encoding_bitrate += (sum_bitrate / rc.window_size) * (sum_bitrate / rc.window_size); sum_bitrate = 0.0; } } // Second shifted window. if (frame_cnt > rc.window_size + rc.window_size / 2) { sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate; if (frame_cnt > 2 * rc.window_size && frame_cnt % rc.window_size == 0) { rc.window_count += 1; rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size; rc.variance_st_encoding_bitrate += (sum_bitrate2 / rc.window_size) * (sum_bitrate2 / rc.window_size); sum_bitrate2 = 0.0; } } break; default: break; } } ++frame_cnt; pts += frame_duration; } fclose(infile); printout_rate_control_summary(&rc, &cfg, frame_cnt); printf("\n"); printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n", frame_cnt, 1000 * (float)cx_time / (double)(frame_cnt * 1000000), 1000000 * (double)frame_cnt / (double)cx_time); if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec"); // Try to rewrite the output file headers with the actual frame count. for (i = 0; i < cfg.ts_number_layers; ++i) vpx_video_writer_close(outfile[i]); vpx_img_free(&raw); return EXIT_SUCCESS; }
int main(int argc, const char **argv) { AppInput app_input = {0}; VpxVideoWriter *writer = NULL; VpxVideoInfo info = {0}; vpx_codec_ctx_t codec; vpx_codec_enc_cfg_t enc_cfg; SvcContext svc_ctx; uint32_t i; uint32_t frame_cnt = 0; vpx_image_t raw; vpx_codec_err_t res; int pts = 0; /* PTS starts at 0 */ int frame_duration = 1; /* 1 timebase tick per frame */ FILE *infile = NULL; memset(&svc_ctx, 0, sizeof(svc_ctx)); svc_ctx.log_print = 1; exec_name = argv[0]; parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg); // Allocate image buffer if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32)) die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h); if (!(infile = fopen(app_input.input_filename, "rb"))) die("Failed to open %s for reading\n", app_input.input_filename); // Initialize codec if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) != VPX_CODEC_OK) die("Failed to initialize encoder\n"); info.codec_fourcc = VP9_FOURCC; info.time_base.numerator = enc_cfg.g_timebase.num; info.time_base.denominator = enc_cfg.g_timebase.den; if (vpx_svc_get_layer_resolution(&svc_ctx, svc_ctx.spatial_layers - 1, (unsigned int *)&info.frame_width, (unsigned int *)&info.frame_height) != VPX_CODEC_OK) { die("Failed to get output resolution"); } writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF, &info); if (!writer) die("Failed to open %s for writing\n", app_input.output_filename); // skip initial frames for (i = 0; i < app_input.frames_to_skip; ++i) vpx_img_read(&raw, infile); // Encode frames while (frame_cnt < app_input.frames_to_code) { if (!vpx_img_read(&raw, infile)) break; res = vpx_svc_encode(&svc_ctx, &codec, &raw, pts, frame_duration, VPX_DL_REALTIME); printf("%s", vpx_svc_get_message(&svc_ctx)); if (res != VPX_CODEC_OK) { die_codec(&codec, "Failed to encode frame"); } if (vpx_svc_get_frame_size(&svc_ctx) > 0) { vpx_video_writer_write_frame(writer, vpx_svc_get_buffer(&svc_ctx), vpx_svc_get_frame_size(&svc_ctx), pts); } ++frame_cnt; pts += frame_duration; } printf("Processed %d frames\n", frame_cnt); fclose(infile); if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec"); vpx_video_writer_close(writer); vpx_img_free(&raw); // display average size, psnr printf("%s", vpx_svc_dump_statistics(&svc_ctx)); vpx_svc_release(&svc_ctx); return EXIT_SUCCESS; }
int main(int argc, const char **argv) { AppInput app_input = {0}; VpxVideoWriter *writer = NULL; VpxVideoInfo info = {0}; vpx_codec_ctx_t codec; vpx_codec_enc_cfg_t enc_cfg; SvcContext svc_ctx; uint32_t i; uint32_t frame_cnt = 0; vpx_image_t raw; vpx_codec_err_t res; int pts = 0; /* PTS starts at 0 */ int frame_duration = 1; /* 1 timebase tick per frame */ FILE *infile = NULL; int end_of_stream = 0; int frames_received = 0; memset(&svc_ctx, 0, sizeof(svc_ctx)); svc_ctx.log_print = 1; exec_name = argv[0]; parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg); // Allocate image buffer if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32)) die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h); if (!(infile = fopen(app_input.input_filename, "rb"))) die("Failed to open %s for reading\n", app_input.input_filename); // Initialize codec if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) != VPX_CODEC_OK) die("Failed to initialize encoder\n"); info.codec_fourcc = VP9_FOURCC; info.time_base.numerator = enc_cfg.g_timebase.num; info.time_base.denominator = enc_cfg.g_timebase.den; if (vpx_svc_get_layer_resolution(&svc_ctx, svc_ctx.spatial_layers - 1, (unsigned int *)&info.frame_width, (unsigned int *)&info.frame_height) != VPX_CODEC_OK) { die("Failed to get output resolution"); } if (!(app_input.passes == 2 && app_input.pass == 1)) { // We don't save the bitstream for the 1st pass on two pass rate control writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF, &info); if (!writer) die("Failed to open %s for writing\n", app_input.output_filename); } // skip initial frames for (i = 0; i < app_input.frames_to_skip; ++i) vpx_img_read(&raw, infile); // Encode frames while (!end_of_stream) { vpx_codec_iter_t iter = NULL; const vpx_codec_cx_pkt_t *cx_pkt; if (frame_cnt >= app_input.frames_to_code || !vpx_img_read(&raw, infile)) { // We need one extra vpx_svc_encode call at end of stream to flush // encoder and get remaining data end_of_stream = 1; } res = vpx_svc_encode(&svc_ctx, &codec, (end_of_stream ? NULL : &raw), pts, frame_duration, VPX_DL_GOOD_QUALITY); printf("%s", vpx_svc_get_message(&svc_ctx)); if (res != VPX_CODEC_OK) { die_codec(&codec, "Failed to encode frame"); } while ((cx_pkt = vpx_codec_get_cx_data(&codec, &iter)) != NULL) { switch (cx_pkt->kind) { case VPX_CODEC_CX_FRAME_PKT: { if (cx_pkt->data.frame.sz > 0) vpx_video_writer_write_frame(writer, cx_pkt->data.frame.buf, cx_pkt->data.frame.sz, cx_pkt->data.frame.pts); printf("SVC frame: %d, kf: %d, size: %d, pts: %d\n", frames_received, !!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY), (int)cx_pkt->data.frame.sz, (int)cx_pkt->data.frame.pts); ++frames_received; break; } case VPX_CODEC_STATS_PKT: { stats_write(&app_input.rc_stats, cx_pkt->data.twopass_stats.buf, cx_pkt->data.twopass_stats.sz); break; } default: { break; } } } if (!end_of_stream) { ++frame_cnt; pts += frame_duration; } } printf("Processed %d frames\n", frame_cnt); fclose(infile); if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec"); if (app_input.passes == 2) stats_close(&app_input.rc_stats, 1); if (writer) { vpx_video_writer_close(writer); } vpx_img_free(&raw); // display average size, psnr printf("%s", vpx_svc_dump_statistics(&svc_ctx)); vpx_svc_release(&svc_ctx); return EXIT_SUCCESS; }