WebM Codec SDK
vpx_temporal_svc_encoder
1/*
2 * Copyright (c) 2012 The WebM project authors. All Rights Reserved.
3 *
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.
9 */
10
11// This is an example demonstrating how to implement a multi-layer VPx
12// encoding scheme based on temporal scalability for video applications
13// that benefit from a scalable bitstream.
14
15#include <assert.h>
16#include <math.h>
17#include <stdio.h>
18#include <stdlib.h>
19#include <string.h>
20
21#include "./vpx_config.h"
22#include "../vpx_ports/vpx_timer.h"
23#include "vpx/vp8cx.h"
24#include "vpx/vpx_encoder.h"
25
26#include "../tools_common.h"
27#include "../video_writer.h"
28
29#define VP8_ROI_MAP 0
30
31static const char *exec_name;
32
33void usage_exit(void) { exit(EXIT_FAILURE); }
34
35// Denoiser states for vp8, for temporal denoising.
36enum denoiserStateVp8 {
37 kVp8DenoiserOff,
38 kVp8DenoiserOnYOnly,
39 kVp8DenoiserOnYUV,
40 kVp8DenoiserOnYUVAggressive,
41 kVp8DenoiserOnAdaptive
42};
43
44// Denoiser states for vp9, for temporal denoising.
45enum denoiserStateVp9 {
46 kVp9DenoiserOff,
47 kVp9DenoiserOnYOnly,
48 // For SVC: denoise the top two spatial layers.
49 kVp9DenoiserOnYTwoSpatialLayers
50};
51
52static int mode_to_num_layers[13] = { 1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3, 3 };
53
54// For rate control encoding stats.
55struct RateControlMetrics {
56 // Number of input frames per layer.
57 int layer_input_frames[VPX_TS_MAX_LAYERS];
58 // Total (cumulative) number of encoded frames per layer.
59 int layer_tot_enc_frames[VPX_TS_MAX_LAYERS];
60 // Number of encoded non-key frames per layer.
61 int layer_enc_frames[VPX_TS_MAX_LAYERS];
62 // Framerate per layer layer (cumulative).
63 double layer_framerate[VPX_TS_MAX_LAYERS];
64 // Target average frame size per layer (per-frame-bandwidth per layer).
65 double layer_pfb[VPX_TS_MAX_LAYERS];
66 // Actual average frame size per layer.
67 double layer_avg_frame_size[VPX_TS_MAX_LAYERS];
68 // Average rate mismatch per layer (|target - actual| / target).
69 double layer_avg_rate_mismatch[VPX_TS_MAX_LAYERS];
70 // Actual encoding bitrate per layer (cumulative).
71 double layer_encoding_bitrate[VPX_TS_MAX_LAYERS];
72 // Average of the short-time encoder actual bitrate.
73 // TODO(marpan): Should we add these short-time stats for each layer?
74 double avg_st_encoding_bitrate;
75 // Variance of the short-time encoder actual bitrate.
76 double variance_st_encoding_bitrate;
77 // Window (number of frames) for computing short-timee encoding bitrate.
78 int window_size;
79 // Number of window measurements.
80 int window_count;
81 int layer_target_bitrate[VPX_MAX_LAYERS];
82};
83
84// Note: these rate control metrics assume only 1 key frame in the
85// sequence (i.e., first frame only). So for temporal pattern# 7
86// (which has key frame for every frame on base layer), the metrics
87// computation will be off/wrong.
88// TODO(marpan): Update these metrics to account for multiple key frames
89// in the stream.
90static void set_rate_control_metrics(struct RateControlMetrics *rc,
92 unsigned int i = 0;
93 // Set the layer (cumulative) framerate and the target layer (non-cumulative)
94 // per-frame-bandwidth, for the rate control encoding stats below.
95 const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
96 rc->layer_framerate[0] = framerate / cfg->ts_rate_decimator[0];
97 rc->layer_pfb[0] =
98 1000.0 * rc->layer_target_bitrate[0] / rc->layer_framerate[0];
99 for (i = 0; i < cfg->ts_number_layers; ++i) {
100 if (i > 0) {
101 rc->layer_framerate[i] = framerate / cfg->ts_rate_decimator[i];
102 rc->layer_pfb[i] = 1000.0 * (rc->layer_target_bitrate[i] -
103 rc->layer_target_bitrate[i - 1]) /
104 (rc->layer_framerate[i] - rc->layer_framerate[i - 1]);
105 }
106 rc->layer_input_frames[i] = 0;
107 rc->layer_enc_frames[i] = 0;
108 rc->layer_tot_enc_frames[i] = 0;
109 rc->layer_encoding_bitrate[i] = 0.0;
110 rc->layer_avg_frame_size[i] = 0.0;
111 rc->layer_avg_rate_mismatch[i] = 0.0;
112 }
113 rc->window_count = 0;
114 rc->window_size = 15;
115 rc->avg_st_encoding_bitrate = 0.0;
116 rc->variance_st_encoding_bitrate = 0.0;
117}
118
119static void printout_rate_control_summary(struct RateControlMetrics *rc,
121 int frame_cnt) {
122 unsigned int i = 0;
123 int tot_num_frames = 0;
124 double perc_fluctuation = 0.0;
125 printf("Total number of processed frames: %d\n\n", frame_cnt - 1);
126 printf("Rate control layer stats for %d layer(s):\n\n",
127 cfg->ts_number_layers);
128 for (i = 0; i < cfg->ts_number_layers; ++i) {
129 const int num_dropped =
130 (i > 0) ? (rc->layer_input_frames[i] - rc->layer_enc_frames[i])
131 : (rc->layer_input_frames[i] - rc->layer_enc_frames[i] - 1);
132 tot_num_frames += rc->layer_input_frames[i];
133 rc->layer_encoding_bitrate[i] = 0.001 * rc->layer_framerate[i] *
134 rc->layer_encoding_bitrate[i] /
135 tot_num_frames;
136 rc->layer_avg_frame_size[i] =
137 rc->layer_avg_frame_size[i] / rc->layer_enc_frames[i];
138 rc->layer_avg_rate_mismatch[i] =
139 100.0 * rc->layer_avg_rate_mismatch[i] / rc->layer_enc_frames[i];
140 printf("For layer#: %d \n", i);
141 printf("Bitrate (target vs actual): %d %f \n", rc->layer_target_bitrate[i],
142 rc->layer_encoding_bitrate[i]);
143 printf("Average frame size (target vs actual): %f %f \n", rc->layer_pfb[i],
144 rc->layer_avg_frame_size[i]);
145 printf("Average rate_mismatch: %f \n", rc->layer_avg_rate_mismatch[i]);
146 printf(
147 "Number of input frames, encoded (non-key) frames, "
148 "and perc dropped frames: %d %d %f \n",
149 rc->layer_input_frames[i], rc->layer_enc_frames[i],
150 100.0 * num_dropped / rc->layer_input_frames[i]);
151 printf("\n");
152 }
153 rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
154 rc->variance_st_encoding_bitrate =
155 rc->variance_st_encoding_bitrate / rc->window_count -
156 (rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
157 perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
158 rc->avg_st_encoding_bitrate;
159 printf("Short-time stats, for window of %d frames: \n", rc->window_size);
160 printf("Average, rms-variance, and percent-fluct: %f %f %f \n",
161 rc->avg_st_encoding_bitrate, sqrt(rc->variance_st_encoding_bitrate),
162 perc_fluctuation);
163 if ((frame_cnt - 1) != tot_num_frames)
164 die("Error: Number of input frames not equal to output! \n");
165}
166
167#if VP8_ROI_MAP
168static void vp8_set_roi_map(vpx_codec_enc_cfg_t *cfg, vpx_roi_map_t *roi) {
169 unsigned int i, j;
170 memset(roi, 0, sizeof(*roi));
171
172 // ROI is based on the segments (4 for vp8, 8 for vp9), smallest unit for
173 // segment is 16x16 for vp8, 8x8 for vp9.
174 roi->rows = (cfg->g_h + 15) / 16;
175 roi->cols = (cfg->g_w + 15) / 16;
176
177 // Applies delta QP on the segment blocks, varies from -63 to 63.
178 // Setting to negative means lower QP (better quality).
179 // Below we set delta_q to the extreme (-63) to show strong effect.
180 roi->delta_q[0] = 0;
181 roi->delta_q[1] = -63;
182 roi->delta_q[2] = 0;
183 roi->delta_q[3] = 0;
184
185 // Applies delta loopfilter strength on the segment blocks, varies from -63 to
186 // 63. Setting to positive means stronger loopfilter.
187 roi->delta_lf[0] = 0;
188 roi->delta_lf[1] = 0;
189 roi->delta_lf[2] = 0;
190 roi->delta_lf[3] = 0;
191
192 // Applies skip encoding threshold on the segment blocks, varies from 0 to
193 // UINT_MAX. Larger value means more skipping of encoding is possible.
194 // This skip threshold only applies on delta frames.
195 roi->static_threshold[0] = 0;
196 roi->static_threshold[1] = 0;
197 roi->static_threshold[2] = 0;
198 roi->static_threshold[3] = 0;
199
200 // Use 2 states: 1 is center square, 0 is the rest.
201 roi->roi_map =
202 (uint8_t *)calloc(roi->rows * roi->cols, sizeof(*roi->roi_map));
203 for (i = 0; i < roi->rows; ++i) {
204 for (j = 0; j < roi->cols; ++j) {
205 if (i > (roi->rows >> 2) && i < ((roi->rows * 3) >> 2) &&
206 j > (roi->cols >> 2) && j < ((roi->cols * 3) >> 2)) {
207 roi->roi_map[i * roi->cols + j] = 1;
208 }
209 }
210 }
211}
212#endif
213
214// Temporal scaling parameters:
215// NOTE: The 3 prediction frames cannot be used interchangeably due to
216// differences in the way they are handled throughout the code. The
217// frames should be allocated to layers in the order LAST, GF, ARF.
218// Other combinations work, but may produce slightly inferior results.
219static void set_temporal_layer_pattern(int layering_mode,
221 int *layer_flags,
222 int *flag_periodicity) {
223 switch (layering_mode) {
224 case 0: {
225 // 1-layer.
226 int ids[1] = { 0 };
227 cfg->ts_periodicity = 1;
228 *flag_periodicity = 1;
229 cfg->ts_number_layers = 1;
230 cfg->ts_rate_decimator[0] = 1;
231 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
232 // Update L only.
233 layer_flags[0] =
235 break;
236 }
237 case 1: {
238 // 2-layers, 2-frame period.
239 int ids[2] = { 0, 1 };
240 cfg->ts_periodicity = 2;
241 *flag_periodicity = 2;
242 cfg->ts_number_layers = 2;
243 cfg->ts_rate_decimator[0] = 2;
244 cfg->ts_rate_decimator[1] = 1;
245 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
246#if 1
247 // 0=L, 1=GF, Intra-layer prediction enabled.
248 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
251 layer_flags[1] =
253#else
254 // 0=L, 1=GF, Intra-layer prediction disabled.
255 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
260#endif
261 break;
262 }
263 case 2: {
264 // 2-layers, 3-frame period.
265 int ids[3] = { 0, 1, 1 };
266 cfg->ts_periodicity = 3;
267 *flag_periodicity = 3;
268 cfg->ts_number_layers = 2;
269 cfg->ts_rate_decimator[0] = 3;
270 cfg->ts_rate_decimator[1] = 1;
271 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
272 // 0=L, 1=GF, Intra-layer prediction enabled.
273 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
276 layer_flags[1] = layer_flags[2] =
279 break;
280 }
281 case 3: {
282 // 3-layers, 6-frame period.
283 int ids[6] = { 0, 2, 2, 1, 2, 2 };
284 cfg->ts_periodicity = 6;
285 *flag_periodicity = 6;
286 cfg->ts_number_layers = 3;
287 cfg->ts_rate_decimator[0] = 6;
288 cfg->ts_rate_decimator[1] = 3;
289 cfg->ts_rate_decimator[2] = 1;
290 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
291 // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
292 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
295 layer_flags[3] =
297 layer_flags[1] = layer_flags[2] = layer_flags[4] = layer_flags[5] =
299 break;
300 }
301 case 4: {
302 // 3-layers, 4-frame period.
303 int ids[4] = { 0, 2, 1, 2 };
304 cfg->ts_periodicity = 4;
305 *flag_periodicity = 4;
306 cfg->ts_number_layers = 3;
307 cfg->ts_rate_decimator[0] = 4;
308 cfg->ts_rate_decimator[1] = 2;
309 cfg->ts_rate_decimator[2] = 1;
310 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
311 // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
312 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
315 layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
317 layer_flags[1] = layer_flags[3] =
320 break;
321 }
322 case 5: {
323 // 3-layers, 4-frame period.
324 int ids[4] = { 0, 2, 1, 2 };
325 cfg->ts_periodicity = 4;
326 *flag_periodicity = 4;
327 cfg->ts_number_layers = 3;
328 cfg->ts_rate_decimator[0] = 4;
329 cfg->ts_rate_decimator[1] = 2;
330 cfg->ts_rate_decimator[2] = 1;
331 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
332 // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled in layer 1, disabled
333 // in layer 2.
334 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
337 layer_flags[2] =
339 layer_flags[1] = layer_flags[3] =
342 break;
343 }
344 case 6: {
345 // 3-layers, 4-frame period.
346 int ids[4] = { 0, 2, 1, 2 };
347 cfg->ts_periodicity = 4;
348 *flag_periodicity = 4;
349 cfg->ts_number_layers = 3;
350 cfg->ts_rate_decimator[0] = 4;
351 cfg->ts_rate_decimator[1] = 2;
352 cfg->ts_rate_decimator[2] = 1;
353 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
354 // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
355 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
358 layer_flags[2] =
360 layer_flags[1] = layer_flags[3] =
362 break;
363 }
364 case 7: {
365 // NOTE: Probably of academic interest only.
366 // 5-layers, 16-frame period.
367 int ids[16] = { 0, 4, 3, 4, 2, 4, 3, 4, 1, 4, 3, 4, 2, 4, 3, 4 };
368 cfg->ts_periodicity = 16;
369 *flag_periodicity = 16;
370 cfg->ts_number_layers = 5;
371 cfg->ts_rate_decimator[0] = 16;
372 cfg->ts_rate_decimator[1] = 8;
373 cfg->ts_rate_decimator[2] = 4;
374 cfg->ts_rate_decimator[3] = 2;
375 cfg->ts_rate_decimator[4] = 1;
376 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
377 layer_flags[0] = VPX_EFLAG_FORCE_KF;
378 layer_flags[1] = layer_flags[3] = layer_flags[5] = layer_flags[7] =
379 layer_flags[9] = layer_flags[11] = layer_flags[13] = layer_flags[15] =
382 layer_flags[2] = layer_flags[6] = layer_flags[10] = layer_flags[14] =
384 layer_flags[4] = layer_flags[12] =
387 break;
388 }
389 case 8: {
390 // 2-layers, with sync point at first frame of layer 1.
391 int ids[2] = { 0, 1 };
392 cfg->ts_periodicity = 2;
393 *flag_periodicity = 8;
394 cfg->ts_number_layers = 2;
395 cfg->ts_rate_decimator[0] = 2;
396 cfg->ts_rate_decimator[1] = 1;
397 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
398 // 0=L, 1=GF.
399 // ARF is used as predictor for all frames, and is only updated on
400 // key frame. Sync point every 8 frames.
401
402 // Layer 0: predict from L and ARF, update L and G.
403 layer_flags[0] =
405 // Layer 1: sync point: predict from L and ARF, and update G.
406 layer_flags[1] =
408 // Layer 0, predict from L and ARF, update L.
409 layer_flags[2] =
411 // Layer 1: predict from L, G and ARF, and update G.
414 // Layer 0.
415 layer_flags[4] = layer_flags[2];
416 // Layer 1.
417 layer_flags[5] = layer_flags[3];
418 // Layer 0.
419 layer_flags[6] = layer_flags[4];
420 // Layer 1.
421 layer_flags[7] = layer_flags[5];
422 break;
423 }
424 case 9: {
425 // 3-layers: Sync points for layer 1 and 2 every 8 frames.
426 int ids[4] = { 0, 2, 1, 2 };
427 cfg->ts_periodicity = 4;
428 *flag_periodicity = 8;
429 cfg->ts_number_layers = 3;
430 cfg->ts_rate_decimator[0] = 4;
431 cfg->ts_rate_decimator[1] = 2;
432 cfg->ts_rate_decimator[2] = 1;
433 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
434 // 0=L, 1=GF, 2=ARF.
435 layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
438 layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
440 layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
442 layer_flags[3] = layer_flags[5] =
444 layer_flags[4] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
446 layer_flags[6] =
448 layer_flags[7] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
450 break;
451 }
452 case 10: {
453 // 3-layers structure where ARF is used as predictor for all frames,
454 // and is only updated on key frame.
455 // Sync points for layer 1 and 2 every 8 frames.
456
457 int ids[4] = { 0, 2, 1, 2 };
458 cfg->ts_periodicity = 4;
459 *flag_periodicity = 8;
460 cfg->ts_number_layers = 3;
461 cfg->ts_rate_decimator[0] = 4;
462 cfg->ts_rate_decimator[1] = 2;
463 cfg->ts_rate_decimator[2] = 1;
464 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
465 // 0=L, 1=GF, 2=ARF.
466 // Layer 0: predict from L and ARF; update L and G.
467 layer_flags[0] =
469 // Layer 2: sync point: predict from L and ARF; update none.
470 layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
473 // Layer 1: sync point: predict from L and ARF; update G.
474 layer_flags[2] =
476 // Layer 2: predict from L, G, ARF; update none.
477 layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
479 // Layer 0: predict from L and ARF; update L.
480 layer_flags[4] =
482 // Layer 2: predict from L, G, ARF; update none.
483 layer_flags[5] = layer_flags[3];
484 // Layer 1: predict from L, G, ARF; update G.
486 // Layer 2: predict from L, G, ARF; update none.
487 layer_flags[7] = layer_flags[3];
488 break;
489 }
490 case 11: {
491 // 3-layers structure with one reference frame.
492 // This works same as temporal_layering_mode 3.
493 // This was added to compare with vp9_spatial_svc_encoder.
494
495 // 3-layers, 4-frame period.
496 int ids[4] = { 0, 2, 1, 2 };
497 cfg->ts_periodicity = 4;
498 *flag_periodicity = 4;
499 cfg->ts_number_layers = 3;
500 cfg->ts_rate_decimator[0] = 4;
501 cfg->ts_rate_decimator[1] = 2;
502 cfg->ts_rate_decimator[2] = 1;
503 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
504 // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
505 layer_flags[0] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
507 layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
509 layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
513 break;
514 }
515 case 12:
516 default: {
517 // 3-layers structure as in case 10, but no sync/refresh points for
518 // layer 1 and 2.
519 int ids[4] = { 0, 2, 1, 2 };
520 cfg->ts_periodicity = 4;
521 *flag_periodicity = 8;
522 cfg->ts_number_layers = 3;
523 cfg->ts_rate_decimator[0] = 4;
524 cfg->ts_rate_decimator[1] = 2;
525 cfg->ts_rate_decimator[2] = 1;
526 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
527 // 0=L, 1=GF, 2=ARF.
528 // Layer 0: predict from L and ARF; update L.
529 layer_flags[0] =
531 layer_flags[4] = layer_flags[0];
532 // Layer 1: predict from L, G, ARF; update G.
534 layer_flags[6] = layer_flags[2];
535 // Layer 2: predict from L, G, ARF; update none.
536 layer_flags[1] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
538 layer_flags[3] = layer_flags[1];
539 layer_flags[5] = layer_flags[1];
540 layer_flags[7] = layer_flags[1];
541 break;
542 }
543 }
544}
545
546int main(int argc, char **argv) {
547 VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = { NULL };
548 vpx_codec_ctx_t codec;
550 int frame_cnt = 0;
551 vpx_image_t raw;
552 vpx_codec_err_t res;
553 unsigned int width;
554 unsigned int height;
555 uint32_t error_resilient = 0;
556 int speed;
557 int frame_avail;
558 int got_data;
559 int flags = 0;
560 unsigned int i;
561 int pts = 0; // PTS starts at 0.
562 int frame_duration = 1; // 1 timebase tick per frame.
563 int layering_mode = 0;
564 int layer_flags[VPX_TS_MAX_PERIODICITY] = { 0 };
565 int flag_periodicity = 1;
566#if VP8_ROI_MAP
567 vpx_roi_map_t roi;
568#endif
569 vpx_svc_layer_id_t layer_id = { 0, 0 };
570 const VpxInterface *encoder = NULL;
571 FILE *infile = NULL;
572 struct RateControlMetrics rc;
573 int64_t cx_time = 0;
574 const int min_args_base = 13;
575#if CONFIG_VP9_HIGHBITDEPTH
576 vpx_bit_depth_t bit_depth = VPX_BITS_8;
577 int input_bit_depth = 8;
578 const int min_args = min_args_base + 1;
579#else
580 const int min_args = min_args_base;
581#endif // CONFIG_VP9_HIGHBITDEPTH
582 double sum_bitrate = 0.0;
583 double sum_bitrate2 = 0.0;
584 double framerate = 30.0;
585
586 exec_name = argv[0];
587 // Check usage and arguments.
588 if (argc < min_args) {
589#if CONFIG_VP9_HIGHBITDEPTH
590 die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
591 "<rate_num> <rate_den> <speed> <frame_drop_threshold> "
592 "<error_resilient> <threads> <mode> "
593 "<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n",
594 argv[0]);
595#else
596 die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
597 "<rate_num> <rate_den> <speed> <frame_drop_threshold> "
598 "<error_resilient> <threads> <mode> "
599 "<Rate_0> ... <Rate_nlayers-1> \n",
600 argv[0]);
601#endif // CONFIG_VP9_HIGHBITDEPTH
602 }
603
604 encoder = get_vpx_encoder_by_name(argv[3]);
605 if (!encoder) die("Unsupported codec.");
606
607 printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
608
609 width = (unsigned int)strtoul(argv[4], NULL, 0);
610 height = (unsigned int)strtoul(argv[5], NULL, 0);
611 if (width < 16 || width % 2 || height < 16 || height % 2) {
612 die("Invalid resolution: %d x %d", width, height);
613 }
614
615 layering_mode = (int)strtol(argv[12], NULL, 0);
616 if (layering_mode < 0 || layering_mode > 13) {
617 die("Invalid layering mode (0..12) %s", argv[12]);
618 }
619
620 if (argc != min_args + mode_to_num_layers[layering_mode]) {
621 die("Invalid number of arguments");
622 }
623
624#if CONFIG_VP9_HIGHBITDEPTH
625 switch (strtol(argv[argc - 1], NULL, 0)) {
626 case 8:
627 bit_depth = VPX_BITS_8;
628 input_bit_depth = 8;
629 break;
630 case 10:
631 bit_depth = VPX_BITS_10;
632 input_bit_depth = 10;
633 break;
634 case 12:
635 bit_depth = VPX_BITS_12;
636 input_bit_depth = 12;
637 break;
638 default: die("Invalid bit depth (8, 10, 12) %s", argv[argc - 1]);
639 }
640 if (!vpx_img_alloc(
641 &raw, bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016,
642 width, height, 32)) {
643 die("Failed to allocate image", width, height);
644 }
645#else
646 if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
647 die("Failed to allocate image", width, height);
648 }
649#endif // CONFIG_VP9_HIGHBITDEPTH
650
651 // Populate encoder configuration.
652 res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
653 if (res) {
654 printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
655 return EXIT_FAILURE;
656 }
657
658 // Update the default configuration with our settings.
659 cfg.g_w = width;
660 cfg.g_h = height;
661
662#if CONFIG_VP9_HIGHBITDEPTH
663 if (bit_depth != VPX_BITS_8) {
664 cfg.g_bit_depth = bit_depth;
665 cfg.g_input_bit_depth = input_bit_depth;
666 cfg.g_profile = 2;
667 }
668#endif // CONFIG_VP9_HIGHBITDEPTH
669
670 // Timebase format e.g. 30fps: numerator=1, demoninator = 30.
671 cfg.g_timebase.num = (int)strtol(argv[6], NULL, 0);
672 cfg.g_timebase.den = (int)strtol(argv[7], NULL, 0);
673
674 speed = (int)strtol(argv[8], NULL, 0);
675 if (speed < 0) {
676 die("Invalid speed setting: must be positive");
677 }
678
679 for (i = min_args_base;
680 (int)i < min_args_base + mode_to_num_layers[layering_mode]; ++i) {
681 rc.layer_target_bitrate[i - 13] = (int)strtol(argv[i], NULL, 0);
682 if (strncmp(encoder->name, "vp8", 3) == 0)
683 cfg.ts_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
684 else if (strncmp(encoder->name, "vp9", 3) == 0)
685 cfg.layer_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
686 }
687
688 // Real time parameters.
689 cfg.rc_dropframe_thresh = (unsigned int)strtoul(argv[9], NULL, 0);
690 cfg.rc_end_usage = VPX_CBR;
691 cfg.rc_min_quantizer = 2;
692 cfg.rc_max_quantizer = 56;
693 if (strncmp(encoder->name, "vp9", 3) == 0) cfg.rc_max_quantizer = 52;
694 cfg.rc_undershoot_pct = 50;
695 cfg.rc_overshoot_pct = 50;
696 cfg.rc_buf_initial_sz = 600;
697 cfg.rc_buf_optimal_sz = 600;
698 cfg.rc_buf_sz = 1000;
699
700 // Disable dynamic resizing by default.
701 cfg.rc_resize_allowed = 0;
702
703 // Use 1 thread as default.
704 cfg.g_threads = (unsigned int)strtoul(argv[11], NULL, 0);
705
706 error_resilient = (uint32_t)strtoul(argv[10], NULL, 0);
707 if (error_resilient != 0 && error_resilient != 1) {
708 die("Invalid value for error resilient (0, 1): %d.", error_resilient);
709 }
710 // Enable error resilient mode.
711 cfg.g_error_resilient = error_resilient;
712 cfg.g_lag_in_frames = 0;
713 cfg.kf_mode = VPX_KF_AUTO;
714
715 // Disable automatic keyframe placement.
716 cfg.kf_min_dist = cfg.kf_max_dist = 3000;
717
719
720 set_temporal_layer_pattern(layering_mode, &cfg, layer_flags,
721 &flag_periodicity);
722
723 set_rate_control_metrics(&rc, &cfg);
724
725 // Target bandwidth for the whole stream.
726 // Set to layer_target_bitrate for highest layer (total bitrate).
727 cfg.rc_target_bitrate = rc.layer_target_bitrate[cfg.ts_number_layers - 1];
728
729 // Open input file.
730 if (!(infile = fopen(argv[1], "rb"))) {
731 die("Failed to open %s for reading", argv[1]);
732 }
733
734 framerate = cfg.g_timebase.den / cfg.g_timebase.num;
735 // Open an output file for each stream.
736 for (i = 0; i < cfg.ts_number_layers; ++i) {
737 char file_name[PATH_MAX];
738 VpxVideoInfo info;
739 info.codec_fourcc = encoder->fourcc;
740 info.frame_width = cfg.g_w;
741 info.frame_height = cfg.g_h;
742 info.time_base.numerator = cfg.g_timebase.num;
743 info.time_base.denominator = cfg.g_timebase.den;
744
745 snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
746 outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
747 if (!outfile[i]) die("Failed to open %s for writing", file_name);
748
749 assert(outfile[i] != NULL);
750 }
751 // No spatial layers in this encoder.
752 cfg.ss_number_layers = 1;
753
754// Initialize codec.
755#if CONFIG_VP9_HIGHBITDEPTH
757 &codec, encoder->codec_interface(), &cfg,
758 bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
759#else
760 if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
761#endif // CONFIG_VP9_HIGHBITDEPTH
762 die_codec(&codec, "Failed to initialize encoder");
763
764 if (strncmp(encoder->name, "vp8", 3) == 0) {
765 vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
766 vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kVp8DenoiserOff);
769#if VP8_ROI_MAP
770 vp8_set_roi_map(&cfg, &roi);
771 if (vpx_codec_control(&codec, VP8E_SET_ROI_MAP, &roi))
772 die_codec(&codec, "Failed to set ROI map");
773#endif
774
775 } else if (strncmp(encoder->name, "vp9", 3) == 0) {
776 vpx_svc_extra_cfg_t svc_params;
777 memset(&svc_params, 0, sizeof(svc_params));
778 vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
783 vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kVp9DenoiserOff);
787 // TODO(marpan/jianj): There is an issue with row-mt for low resolutons at
788 // high speed settings, disable its use for those cases for now.
789 if (cfg.g_threads > 1 && ((cfg.g_w > 320 && cfg.g_h > 240) || speed < 7))
791 else
793 if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1 : 0))
794 die_codec(&codec, "Failed to set SVC");
795 for (i = 0; i < cfg.ts_number_layers; ++i) {
796 svc_params.max_quantizers[i] = cfg.rc_max_quantizer;
797 svc_params.min_quantizers[i] = cfg.rc_min_quantizer;
798 }
799 svc_params.scaling_factor_num[0] = cfg.g_h;
800 svc_params.scaling_factor_den[0] = cfg.g_h;
801 vpx_codec_control(&codec, VP9E_SET_SVC_PARAMETERS, &svc_params);
802 }
803 if (strncmp(encoder->name, "vp8", 3) == 0) {
805 }
807 // This controls the maximum target size of the key frame.
808 // For generating smaller key frames, use a smaller max_intra_size_pct
809 // value, like 100 or 200.
810 {
811 const int max_intra_size_pct = 1000;
813 max_intra_size_pct);
814 }
815
816 frame_avail = 1;
817 while (frame_avail || got_data) {
818 struct vpx_usec_timer timer;
819 vpx_codec_iter_t iter = NULL;
820 const vpx_codec_cx_pkt_t *pkt;
821 // Update the temporal layer_id. No spatial layers in this test.
822 layer_id.spatial_layer_id = 0;
823 layer_id.temporal_layer_id =
824 cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
825 if (strncmp(encoder->name, "vp9", 3) == 0) {
826 vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
827 } else if (strncmp(encoder->name, "vp8", 3) == 0) {
829 layer_id.temporal_layer_id);
830 }
831 flags = layer_flags[frame_cnt % flag_periodicity];
832 if (layering_mode == 0) flags = 0;
833 frame_avail = vpx_img_read(&raw, infile);
834 if (frame_avail) ++rc.layer_input_frames[layer_id.temporal_layer_id];
835 vpx_usec_timer_start(&timer);
836 if (vpx_codec_encode(&codec, frame_avail ? &raw : NULL, pts, 1, flags,
838 die_codec(&codec, "Failed to encode frame");
839 }
840 vpx_usec_timer_mark(&timer);
841 cx_time += vpx_usec_timer_elapsed(&timer);
842 // Reset KF flag.
843 if (layering_mode != 7) {
844 layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
845 }
846 got_data = 0;
847 while ((pkt = vpx_codec_get_cx_data(&codec, &iter))) {
848 got_data = 1;
849 switch (pkt->kind) {
851 for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
852 i < cfg.ts_number_layers; ++i) {
853 vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
854 pkt->data.frame.sz, pts);
855 ++rc.layer_tot_enc_frames[i];
856 rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
857 // Keep count of rate control stats per layer (for non-key frames).
858 if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
859 !(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
860 rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
861 rc.layer_avg_rate_mismatch[i] +=
862 fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
863 rc.layer_pfb[i];
864 ++rc.layer_enc_frames[i];
865 }
866 }
867 // Update for short-time encoding bitrate states, for moving window
868 // of size rc->window, shifted by rc->window / 2.
869 // Ignore first window segment, due to key frame.
870 if (frame_cnt > rc.window_size) {
871 sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
872 if (frame_cnt % rc.window_size == 0) {
873 rc.window_count += 1;
874 rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
875 rc.variance_st_encoding_bitrate +=
876 (sum_bitrate / rc.window_size) *
877 (sum_bitrate / rc.window_size);
878 sum_bitrate = 0.0;
879 }
880 }
881 // Second shifted window.
882 if (frame_cnt > rc.window_size + rc.window_size / 2) {
883 sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
884 if (frame_cnt > 2 * rc.window_size &&
885 frame_cnt % rc.window_size == 0) {
886 rc.window_count += 1;
887 rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
888 rc.variance_st_encoding_bitrate +=
889 (sum_bitrate2 / rc.window_size) *
890 (sum_bitrate2 / rc.window_size);
891 sum_bitrate2 = 0.0;
892 }
893 }
894 break;
895 default: break;
896 }
897 }
898 ++frame_cnt;
899 pts += frame_duration;
900 }
901 fclose(infile);
902 printout_rate_control_summary(&rc, &cfg, frame_cnt);
903 printf("\n");
904 printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
905 frame_cnt, 1000 * (float)cx_time / (double)(frame_cnt * 1000000),
906 1000000 * (double)frame_cnt / (double)cx_time);
907
908 if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
909
910 // Try to rewrite the output file headers with the actual frame count.
911 for (i = 0; i < cfg.ts_number_layers; ++i) vpx_video_writer_close(outfile[i]);
912
913 vpx_img_free(&raw);
914 return EXIT_SUCCESS;
915}
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
vpx_codec_err_t vpx_codec_destroy(vpx_codec_ctx_t *ctx)
Destroy a codec instance.
const void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:187
const char * vpx_codec_iface_name(vpx_codec_iface_t *iface)
Return the name for a given interface.
enum vpx_bit_depth vpx_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition: vpx_codec.h:404
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:90
@ VPX_BITS_8
Definition: vpx_codec.h:218
@ VPX_BITS_12
Definition: vpx_codec.h:220
@ VPX_BITS_10
Definition: vpx_codec.h:219
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition: vpx_encoder.h:849
#define VPX_TS_MAX_LAYERS
Definition: vpx_encoder.h:40
#define vpx_codec_enc_init(ctx, iface, cfg, flags)
Convenience macro for vpx_codec_enc_init_ver()
Definition: vpx_encoder.h:760
#define VPX_EFLAG_FORCE_KF
Definition: vpx_encoder.h:271
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
#define VPX_TS_MAX_PERIODICITY
Definition: vpx_encoder.h:37
#define VPX_CODEC_USE_HIGHBITDEPTH
Definition: vpx_encoder.h:96
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int reserved)
Get a default configuration.
#define VPX_MAX_LAYERS
Definition: vpx_encoder.h:46
#define VPX_FRAME_IS_KEY
Definition: vpx_encoder.h:122
vpx_codec_err_t vpx_codec_encode(vpx_codec_ctx_t *ctx, const vpx_image_t *img, vpx_codec_pts_t pts, unsigned long duration, vpx_enc_frame_flags_t flags, unsigned long deadline)
Encode a frame.
@ VPX_CODEC_CX_FRAME_PKT
Definition: vpx_encoder.h:153
@ VPX_KF_AUTO
Definition: vpx_encoder.h:259
@ VPX_CBR
Definition: vpx_encoder.h:244
#define VP8_EFLAG_NO_UPD_ARF
Don't update the alternate reference frame.
Definition: vp8cx.h:95
#define VP8_EFLAG_NO_UPD_ENTROPY
Disable entropy update.
Definition: vp8cx.h:116
#define VP8_EFLAG_NO_UPD_LAST
Don't update the last frame.
Definition: vp8cx.h:81
#define VP8_EFLAG_NO_REF_ARF
Don't reference the alternate reference frame.
Definition: vp8cx.h:74
#define VP8_EFLAG_NO_UPD_GF
Don't update the golden frame.
Definition: vp8cx.h:88
#define VP8_EFLAG_NO_REF_GF
Don't reference the golden frame.
Definition: vp8cx.h:66
#define VP8_EFLAG_NO_REF_LAST
Don't reference the last frame.
Definition: vp8cx.h:58
@ VP9E_SET_FRAME_PERIODIC_BOOST
Codec control function to enable/disable periodic Q boost.
Definition: vp8cx.h:407
@ VP9E_SET_SVC_LAYER_ID
Codec control function to set svc layer for spatial and temporal.
Definition: vp8cx.h:441
@ VP8E_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set Max data rate for Intra frames.
Definition: vp8cx.h:251
@ VP8E_SET_ROI_MAP
Codec control function to pass an ROI map to encoder.
Definition: vp8cx.h:130
@ VP9E_SET_AQ_MODE
Codec control function to set adaptive quantization mode.
Definition: vp8cx.h:392
@ VP8E_SET_NOISE_SENSITIVITY
control function to set noise sensitivity
Definition: vp8cx.h:170
@ VP8E_SET_TOKEN_PARTITIONS
Codec control function to set the number of token partitions.
Definition: vp8cx.h:188
@ VP9E_SET_SVC_PARAMETERS
Codec control function to set parameters for SVC.
Definition: vp8cx.h:432
@ VP9E_SET_FRAME_PARALLEL_DECODING
Codec control function to enable frame parallel decoding feature.
Definition: vp8cx.h:379
@ VP8E_SET_GF_CBR_BOOST_PCT
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:589
@ VP9E_SET_TUNE_CONTENT
Codec control function to set content type.
Definition: vp8cx.h:451
@ VP9E_SET_SVC
Codec control function to turn on/off SVC in encoder.
Definition: vp8cx.h:424
@ VP9E_SET_ROW_MT
Codec control function to set row level multi-threading.
Definition: vp8cx.h:558
@ VP8E_SET_CPUUSED
Codec control function to set encoder internal speed settings.
Definition: vp8cx.h:155
@ VP8E_SET_TEMPORAL_LAYER_ID
Codec control function to set the temporal layer id.
Definition: vp8cx.h:298
@ VP9E_SET_TILE_COLUMNS
Codec control function to set number of tile columns.
Definition: vp8cx.h:345
@ VP8E_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static.
Definition: vp8cx.h:182
@ VP8E_SET_SCREEN_CONTENT_MODE
Codec control function to set encoder screen content mode.
Definition: vp8cx.h:306
@ VP9E_SET_NOISE_SENSITIVITY
Codec control function to set noise sensitivity.
Definition: vp8cx.h:415
@ VP9E_SET_GF_CBR_BOOST_PCT
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:287
@ VP9E_TEMPORAL_LAYERING_MODE_BYPASS
Bypass mode. Used when application needs to control temporal layering. This will only work when the n...
Definition: vp8cx.h:628
Codec context structure.
Definition: vpx_codec.h:197
Encoder output packet.
Definition: vpx_encoder.h:170
vpx_codec_frame_flags_t flags
Definition: vpx_encoder.h:180
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:171
struct vpx_codec_cx_pkt::@1::@2 frame
size_t sz
Definition: vpx_encoder.h:175
void * buf
Definition: vpx_encoder.h:174
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition: vpx_encoder.h:279
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:417
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:712
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:624
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:491
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:663
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:643
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:309
unsigned int layer_target_bitrate[12]
Target bitrate for each spatial/temporal layer.
Definition: vpx_encoder.h:703
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:327
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:615
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition: vpx_encoder.h:695
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:365
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:686
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:535
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:318
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:550
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:408
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:357
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:501
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:386
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:457
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:559
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:335
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:568
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:477
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition: vpx_encoder.h:670
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:343
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:520
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:677
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:633
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:299
Image Descriptor.
Definition: vpx_image.h:88
int den
Definition: vpx_encoder.h:231
int num
Definition: vpx_encoder.h:230
vpx region of interest map
Definition: vp8cx.h:645
unsigned int static_threshold[4]
Definition: vp8cx.h:656
unsigned int rows
Definition: vp8cx.h:648
int delta_lf[4]
Definition: vp8cx.h:654
unsigned int cols
Definition: vp8cx.h:649
int delta_q[4]
Definition: vp8cx.h:653
unsigned char * roi_map
Definition: vp8cx.h:647
vp9 svc layer parameters
Definition: vp8cx.h:718
int temporal_layer_id
Definition: vp8cx.h:720
int spatial_layer_id
Definition: vp8cx.h:719
vp9 svc extra configure parameters
Definition: vpx_encoder.h:720
int min_quantizers[12]
Definition: vpx_encoder.h:722
int scaling_factor_num[12]
Definition: vpx_encoder.h:723
int max_quantizers[12]
Definition: vpx_encoder.h:721
int scaling_factor_den[12]
Definition: vpx_encoder.h:724
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
Describes the encoder algorithm interface to applications.
vpx_image_t * vpx_img_alloc(vpx_image_t *img, vpx_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
@ VPX_IMG_FMT_I42016
Definition: vpx_image.h:63
@ VPX_IMG_FMT_I420
Definition: vpx_image.h:55
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.