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 "./y4minput.h"
23 #include "../vpx_ports/vpx_timer.h"
24 #include "vpx/vp8cx.h"
25 #include "vpx/vpx_encoder.h"
26 #include "vpx_ports/bitops.h"
27 
28 #include "../tools_common.h"
29 #include "../video_writer.h"
30 
31 #define ROI_MAP 0
32 
33 #define zero(Dest) memset(&(Dest), 0, sizeof(Dest));
34 
35 static const char *exec_name;
36 
37 void usage_exit(void) { exit(EXIT_FAILURE); }
38 
39 // Denoiser states for vp8, for temporal denoising.
40 enum denoiserStateVp8 {
41  kVp8DenoiserOff,
42  kVp8DenoiserOnYOnly,
43  kVp8DenoiserOnYUV,
44  kVp8DenoiserOnYUVAggressive,
45  kVp8DenoiserOnAdaptive
46 };
47 
48 // Denoiser states for vp9, for temporal denoising.
49 enum denoiserStateVp9 {
50  kVp9DenoiserOff,
51  kVp9DenoiserOnYOnly,
52  // For SVC: denoise the top two spatial layers.
53  kVp9DenoiserOnYTwoSpatialLayers
54 };
55 
56 static int mode_to_num_layers[13] = { 1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3, 3 };
57 
58 // For rate control encoding stats.
59 struct RateControlMetrics {
60  // Number of input frames per layer.
61  int layer_input_frames[VPX_TS_MAX_LAYERS];
62  // Total (cumulative) number of encoded frames per layer.
63  int layer_tot_enc_frames[VPX_TS_MAX_LAYERS];
64  // Number of encoded non-key frames per layer.
65  int layer_enc_frames[VPX_TS_MAX_LAYERS];
66  // Framerate per layer layer (cumulative).
67  double layer_framerate[VPX_TS_MAX_LAYERS];
68  // Target average frame size per layer (per-frame-bandwidth per layer).
69  double layer_pfb[VPX_TS_MAX_LAYERS];
70  // Actual average frame size per layer.
71  double layer_avg_frame_size[VPX_TS_MAX_LAYERS];
72  // Average rate mismatch per layer (|target - actual| / target).
73  double layer_avg_rate_mismatch[VPX_TS_MAX_LAYERS];
74  // Actual encoding bitrate per layer (cumulative).
75  double layer_encoding_bitrate[VPX_TS_MAX_LAYERS];
76  // Average of the short-time encoder actual bitrate.
77  // TODO(marpan): Should we add these short-time stats for each layer?
78  double avg_st_encoding_bitrate;
79  // Variance of the short-time encoder actual bitrate.
80  double variance_st_encoding_bitrate;
81  // Window (number of frames) for computing short-timee encoding bitrate.
82  int window_size;
83  // Number of window measurements.
84  int window_count;
85  int layer_target_bitrate[VPX_MAX_LAYERS];
86 };
87 
88 // Note: these rate control metrics assume only 1 key frame in the
89 // sequence (i.e., first frame only). So for temporal pattern# 7
90 // (which has key frame for every frame on base layer), the metrics
91 // computation will be off/wrong.
92 // TODO(marpan): Update these metrics to account for multiple key frames
93 // in the stream.
94 static void set_rate_control_metrics(struct RateControlMetrics *rc,
95  vpx_codec_enc_cfg_t *cfg) {
96  int i = 0;
97  // Set the layer (cumulative) framerate and the target layer (non-cumulative)
98  // per-frame-bandwidth, for the rate control encoding stats below.
99  const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
100  const int ts_number_layers = cfg->ts_number_layers;
101  rc->layer_framerate[0] = framerate / cfg->ts_rate_decimator[0];
102  rc->layer_pfb[0] =
103  1000.0 * rc->layer_target_bitrate[0] / rc->layer_framerate[0];
104  for (i = 0; i < ts_number_layers; ++i) {
105  if (i > 0) {
106  rc->layer_framerate[i] = framerate / cfg->ts_rate_decimator[i];
107  rc->layer_pfb[i] =
108  1000.0 *
109  (rc->layer_target_bitrate[i] - rc->layer_target_bitrate[i - 1]) /
110  (rc->layer_framerate[i] - rc->layer_framerate[i - 1]);
111  }
112  rc->layer_input_frames[i] = 0;
113  rc->layer_enc_frames[i] = 0;
114  rc->layer_tot_enc_frames[i] = 0;
115  rc->layer_encoding_bitrate[i] = 0.0;
116  rc->layer_avg_frame_size[i] = 0.0;
117  rc->layer_avg_rate_mismatch[i] = 0.0;
118  }
119  rc->window_count = 0;
120  rc->window_size = 15;
121  rc->avg_st_encoding_bitrate = 0.0;
122  rc->variance_st_encoding_bitrate = 0.0;
123  // Target bandwidth for the whole stream.
124  // Set to layer_target_bitrate for highest layer (total bitrate).
125  cfg->rc_target_bitrate = rc->layer_target_bitrate[ts_number_layers - 1];
126 }
127 
128 static void printout_rate_control_summary(struct RateControlMetrics *rc,
129  vpx_codec_enc_cfg_t *cfg,
130  int frame_cnt) {
131  unsigned int i = 0;
132  int tot_num_frames = 0;
133  double perc_fluctuation = 0.0;
134  printf("Total number of processed frames: %d\n\n", frame_cnt - 1);
135  printf("Rate control layer stats for %d layer(s):\n\n",
136  cfg->ts_number_layers);
137  for (i = 0; i < cfg->ts_number_layers; ++i) {
138  const int num_dropped =
139  (i > 0) ? (rc->layer_input_frames[i] - rc->layer_enc_frames[i])
140  : (rc->layer_input_frames[i] - rc->layer_enc_frames[i] - 1);
141  tot_num_frames += rc->layer_input_frames[i];
142  rc->layer_encoding_bitrate[i] = 0.001 * rc->layer_framerate[i] *
143  rc->layer_encoding_bitrate[i] /
144  tot_num_frames;
145  rc->layer_avg_frame_size[i] =
146  rc->layer_avg_frame_size[i] / rc->layer_enc_frames[i];
147  rc->layer_avg_rate_mismatch[i] =
148  100.0 * rc->layer_avg_rate_mismatch[i] / rc->layer_enc_frames[i];
149  printf("For layer#: %d \n", i);
150  printf("Bitrate (target vs actual): %d %f \n", rc->layer_target_bitrate[i],
151  rc->layer_encoding_bitrate[i]);
152  printf("Average frame size (target vs actual): %f %f \n", rc->layer_pfb[i],
153  rc->layer_avg_frame_size[i]);
154  printf("Average rate_mismatch: %f \n", rc->layer_avg_rate_mismatch[i]);
155  printf(
156  "Number of input frames, encoded (non-key) frames, "
157  "and perc dropped frames: %d %d %f \n",
158  rc->layer_input_frames[i], rc->layer_enc_frames[i],
159  100.0 * num_dropped / rc->layer_input_frames[i]);
160  printf("\n");
161  }
162  rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
163  rc->variance_st_encoding_bitrate =
164  rc->variance_st_encoding_bitrate / rc->window_count -
165  (rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
166  perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
167  rc->avg_st_encoding_bitrate;
168  printf("Short-time stats, for window of %d frames: \n", rc->window_size);
169  printf("Average, rms-variance, and percent-fluct: %f %f %f \n",
170  rc->avg_st_encoding_bitrate, sqrt(rc->variance_st_encoding_bitrate),
171  perc_fluctuation);
172  if ((frame_cnt - 1) != tot_num_frames)
173  die("Error: Number of input frames not equal to output! \n");
174 }
175 
176 #if ROI_MAP
177 static void set_roi_map(const char *enc_name, vpx_codec_enc_cfg_t *cfg,
178  vpx_roi_map_t *roi) {
179  unsigned int i, j;
180  int block_size = 0;
181  uint8_t is_vp8 = strncmp(enc_name, "vp8", 3) == 0 ? 1 : 0;
182  uint8_t is_vp9 = strncmp(enc_name, "vp9", 3) == 0 ? 1 : 0;
183  if (!is_vp8 && !is_vp9) {
184  die("unsupported codec.");
185  }
186  zero(*roi);
187 
188  block_size = is_vp9 && !is_vp8 ? 8 : 16;
189 
190  // ROI is based on the segments (4 for vp8, 8 for vp9), smallest unit for
191  // segment is 16x16 for vp8, 8x8 for vp9.
192  roi->rows = (cfg->g_h + block_size - 1) / block_size;
193  roi->cols = (cfg->g_w + block_size - 1) / block_size;
194 
195  // Applies delta QP on the segment blocks, varies from -63 to 63.
196  // Setting to negative means lower QP (better quality).
197  // Below we set delta_q to the extreme (-63) to show strong effect.
198  // VP8 uses the first 4 segments. VP9 uses all 8 segments.
199  zero(roi->delta_q);
200  roi->delta_q[1] = -63;
201 
202  // Applies delta loopfilter strength on the segment blocks, varies from -63 to
203  // 63. Setting to positive means stronger loopfilter. VP8 uses the first 4
204  // segments. VP9 uses all 8 segments.
205  zero(roi->delta_lf);
206 
207  if (is_vp8) {
208  // Applies skip encoding threshold on the segment blocks, varies from 0 to
209  // UINT_MAX. Larger value means more skipping of encoding is possible.
210  // This skip threshold only applies on delta frames.
211  zero(roi->static_threshold);
212  }
213 
214  if (is_vp9) {
215  // Apply skip segment. Setting to 1 means this block will be copied from
216  // previous frame.
217  zero(roi->skip);
218  }
219 
220  if (is_vp9) {
221  // Apply ref frame segment.
222  // -1 : Do not apply this segment.
223  // 0 : Froce using intra.
224  // 1 : Force using last.
225  // 2 : Force using golden.
226  // 3 : Force using alfref but not used in non-rd pickmode for 0 lag.
227  memset(roi->ref_frame, -1, sizeof(roi->ref_frame));
228  roi->ref_frame[1] = 1;
229  }
230 
231  // Use 2 states: 1 is center square, 0 is the rest.
232  roi->roi_map =
233  (uint8_t *)calloc(roi->rows * roi->cols, sizeof(*roi->roi_map));
234  for (i = 0; i < roi->rows; ++i) {
235  for (j = 0; j < roi->cols; ++j) {
236  if (i > (roi->rows >> 2) && i < ((roi->rows * 3) >> 2) &&
237  j > (roi->cols >> 2) && j < ((roi->cols * 3) >> 2)) {
238  roi->roi_map[i * roi->cols + j] = 1;
239  }
240  }
241  }
242 }
243 #endif
244 
245 // Temporal scaling parameters:
246 // NOTE: The 3 prediction frames cannot be used interchangeably due to
247 // differences in the way they are handled throughout the code. The
248 // frames should be allocated to layers in the order LAST, GF, ARF.
249 // Other combinations work, but may produce slightly inferior results.
250 static void set_temporal_layer_pattern(int layering_mode,
251  vpx_codec_enc_cfg_t *cfg,
252  int *layer_flags,
253  int *flag_periodicity) {
254  switch (layering_mode) {
255  case 0: {
256  // 1-layer.
257  int ids[1] = { 0 };
258  cfg->ts_periodicity = 1;
259  *flag_periodicity = 1;
260  cfg->ts_number_layers = 1;
261  cfg->ts_rate_decimator[0] = 1;
262  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
263  // Update L only.
264  layer_flags[0] =
266  break;
267  }
268  case 1: {
269  // 2-layers, 2-frame period.
270  int ids[2] = { 0, 1 };
271  cfg->ts_periodicity = 2;
272  *flag_periodicity = 2;
273  cfg->ts_number_layers = 2;
274  cfg->ts_rate_decimator[0] = 2;
275  cfg->ts_rate_decimator[1] = 1;
276  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
277 #if 1
278  // 0=L, 1=GF, Intra-layer prediction enabled.
279  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
282  layer_flags[1] =
284 #else
285  // 0=L, 1=GF, Intra-layer prediction disabled.
286  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
289  layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
291 #endif
292  break;
293  }
294  case 2: {
295  // 2-layers, 3-frame period.
296  int ids[3] = { 0, 1, 1 };
297  cfg->ts_periodicity = 3;
298  *flag_periodicity = 3;
299  cfg->ts_number_layers = 2;
300  cfg->ts_rate_decimator[0] = 3;
301  cfg->ts_rate_decimator[1] = 1;
302  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
303  // 0=L, 1=GF, Intra-layer prediction enabled.
304  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
307  layer_flags[1] = layer_flags[2] =
310  break;
311  }
312  case 3: {
313  // 3-layers, 6-frame period.
314  int ids[6] = { 0, 2, 2, 1, 2, 2 };
315  cfg->ts_periodicity = 6;
316  *flag_periodicity = 6;
317  cfg->ts_number_layers = 3;
318  cfg->ts_rate_decimator[0] = 6;
319  cfg->ts_rate_decimator[1] = 3;
320  cfg->ts_rate_decimator[2] = 1;
321  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
322  // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
323  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
326  layer_flags[3] =
328  layer_flags[1] = layer_flags[2] = layer_flags[4] = layer_flags[5] =
330  break;
331  }
332  case 4: {
333  // 3-layers, 4-frame period.
334  int ids[4] = { 0, 2, 1, 2 };
335  cfg->ts_periodicity = 4;
336  *flag_periodicity = 4;
337  cfg->ts_number_layers = 3;
338  cfg->ts_rate_decimator[0] = 4;
339  cfg->ts_rate_decimator[1] = 2;
340  cfg->ts_rate_decimator[2] = 1;
341  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
342  // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
343  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
346  layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
348  layer_flags[1] = layer_flags[3] =
351  break;
352  }
353  case 5: {
354  // 3-layers, 4-frame period.
355  int ids[4] = { 0, 2, 1, 2 };
356  cfg->ts_periodicity = 4;
357  *flag_periodicity = 4;
358  cfg->ts_number_layers = 3;
359  cfg->ts_rate_decimator[0] = 4;
360  cfg->ts_rate_decimator[1] = 2;
361  cfg->ts_rate_decimator[2] = 1;
362  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
363  // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled in layer 1, disabled
364  // in layer 2.
365  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
368  layer_flags[2] =
370  layer_flags[1] = layer_flags[3] =
373  break;
374  }
375  case 6: {
376  // 3-layers, 4-frame period.
377  int ids[4] = { 0, 2, 1, 2 };
378  cfg->ts_periodicity = 4;
379  *flag_periodicity = 4;
380  cfg->ts_number_layers = 3;
381  cfg->ts_rate_decimator[0] = 4;
382  cfg->ts_rate_decimator[1] = 2;
383  cfg->ts_rate_decimator[2] = 1;
384  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
385  // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
386  layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
389  layer_flags[2] =
391  layer_flags[1] = layer_flags[3] =
393  break;
394  }
395  case 7: {
396  // NOTE: Probably of academic interest only.
397  // 5-layers, 16-frame period.
398  int ids[16] = { 0, 4, 3, 4, 2, 4, 3, 4, 1, 4, 3, 4, 2, 4, 3, 4 };
399  cfg->ts_periodicity = 16;
400  *flag_periodicity = 16;
401  cfg->ts_number_layers = 5;
402  cfg->ts_rate_decimator[0] = 16;
403  cfg->ts_rate_decimator[1] = 8;
404  cfg->ts_rate_decimator[2] = 4;
405  cfg->ts_rate_decimator[3] = 2;
406  cfg->ts_rate_decimator[4] = 1;
407  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
408  layer_flags[0] = VPX_EFLAG_FORCE_KF;
409  layer_flags[1] = layer_flags[3] = layer_flags[5] = layer_flags[7] =
410  layer_flags[9] = layer_flags[11] = layer_flags[13] = layer_flags[15] =
413  layer_flags[2] = layer_flags[6] = layer_flags[10] = layer_flags[14] =
415  layer_flags[4] = layer_flags[12] =
417  layer_flags[8] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF;
418  break;
419  }
420  case 8: {
421  // 2-layers, with sync point at first frame of layer 1.
422  int ids[2] = { 0, 1 };
423  cfg->ts_periodicity = 2;
424  *flag_periodicity = 8;
425  cfg->ts_number_layers = 2;
426  cfg->ts_rate_decimator[0] = 2;
427  cfg->ts_rate_decimator[1] = 1;
428  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
429  // 0=L, 1=GF.
430  // ARF is used as predictor for all frames, and is only updated on
431  // key frame. Sync point every 8 frames.
432 
433  // Layer 0: predict from L and ARF, update L and G.
434  layer_flags[0] =
436  // Layer 1: sync point: predict from L and ARF, and update G.
437  layer_flags[1] =
439  // Layer 0, predict from L and ARF, update L.
440  layer_flags[2] =
442  // Layer 1: predict from L, G and ARF, and update G.
443  layer_flags[3] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
445  // Layer 0.
446  layer_flags[4] = layer_flags[2];
447  // Layer 1.
448  layer_flags[5] = layer_flags[3];
449  // Layer 0.
450  layer_flags[6] = layer_flags[4];
451  // Layer 1.
452  layer_flags[7] = layer_flags[5];
453  break;
454  }
455  case 9: {
456  // 3-layers: Sync points for layer 1 and 2 every 8 frames.
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_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
469  layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
471  layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
473  layer_flags[3] = layer_flags[5] =
475  layer_flags[4] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
477  layer_flags[6] =
479  layer_flags[7] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
481  break;
482  }
483  case 10: {
484  // 3-layers structure where ARF is used as predictor for all frames,
485  // and is only updated on key frame.
486  // Sync points for layer 1 and 2 every 8 frames.
487 
488  int ids[4] = { 0, 2, 1, 2 };
489  cfg->ts_periodicity = 4;
490  *flag_periodicity = 8;
491  cfg->ts_number_layers = 3;
492  cfg->ts_rate_decimator[0] = 4;
493  cfg->ts_rate_decimator[1] = 2;
494  cfg->ts_rate_decimator[2] = 1;
495  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
496  // 0=L, 1=GF, 2=ARF.
497  // Layer 0: predict from L and ARF; update L and G.
498  layer_flags[0] =
500  // Layer 2: sync point: predict from L and ARF; update none.
501  layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
504  // Layer 1: sync point: predict from L and ARF; update G.
505  layer_flags[2] =
507  // Layer 2: predict from L, G, ARF; update none.
508  layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
510  // Layer 0: predict from L and ARF; update L.
511  layer_flags[4] =
513  // Layer 2: predict from L, G, ARF; update none.
514  layer_flags[5] = layer_flags[3];
515  // Layer 1: predict from L, G, ARF; update G.
516  layer_flags[6] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
517  // Layer 2: predict from L, G, ARF; update none.
518  layer_flags[7] = layer_flags[3];
519  break;
520  }
521  case 11: {
522  // 3-layers structure with one reference frame.
523  // This works same as temporal_layering_mode 3.
524  // This was added to compare with vp9_spatial_svc_encoder.
525 
526  // 3-layers, 4-frame period.
527  int ids[4] = { 0, 2, 1, 2 };
528  cfg->ts_periodicity = 4;
529  *flag_periodicity = 4;
530  cfg->ts_number_layers = 3;
531  cfg->ts_rate_decimator[0] = 4;
532  cfg->ts_rate_decimator[1] = 2;
533  cfg->ts_rate_decimator[2] = 1;
534  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
535  // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
536  layer_flags[0] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
538  layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
540  layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
542  layer_flags[3] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_ARF |
544  break;
545  }
546  case 12:
547  default: {
548  // 3-layers structure as in case 10, but no sync/refresh points for
549  // layer 1 and 2.
550  int ids[4] = { 0, 2, 1, 2 };
551  cfg->ts_periodicity = 4;
552  *flag_periodicity = 8;
553  cfg->ts_number_layers = 3;
554  cfg->ts_rate_decimator[0] = 4;
555  cfg->ts_rate_decimator[1] = 2;
556  cfg->ts_rate_decimator[2] = 1;
557  memcpy(cfg->ts_layer_id, ids, sizeof(ids));
558  // 0=L, 1=GF, 2=ARF.
559  // Layer 0: predict from L and ARF; update L.
560  layer_flags[0] =
562  layer_flags[4] = layer_flags[0];
563  // Layer 1: predict from L, G, ARF; update G.
564  layer_flags[2] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
565  layer_flags[6] = layer_flags[2];
566  // Layer 2: predict from L, G, ARF; update none.
567  layer_flags[1] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
569  layer_flags[3] = layer_flags[1];
570  layer_flags[5] = layer_flags[1];
571  layer_flags[7] = layer_flags[1];
572  break;
573  }
574  }
575 }
576 
577 int main(int argc, char **argv) {
578  VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = { NULL };
579  vpx_codec_ctx_t codec;
581  int frame_cnt = 0;
582  vpx_image_t raw;
583  vpx_codec_err_t res;
584  unsigned int width;
585  unsigned int height;
586  uint32_t error_resilient = 0;
587  int speed;
588  int frame_avail;
589  int got_data;
590  int flags = 0;
591  unsigned int i;
592  int pts = 0; // PTS starts at 0.
593  int frame_duration = 1; // 1 timebase tick per frame.
594  int layering_mode = 0;
595  int layer_flags[VPX_TS_MAX_PERIODICITY] = { 0 };
596  int flag_periodicity = 1;
597 #if ROI_MAP
598  vpx_roi_map_t roi;
599 #endif
600  vpx_svc_layer_id_t layer_id;
601  const VpxInterface *encoder = NULL;
602  struct VpxInputContext input_ctx;
603  struct RateControlMetrics rc;
604  int64_t cx_time = 0;
605  const int min_args_base = 13;
606 #if CONFIG_VP9_HIGHBITDEPTH
607  vpx_bit_depth_t bit_depth = VPX_BITS_8;
608  int input_bit_depth = 8;
609  const int min_args = min_args_base + 1;
610 #else
611  const int min_args = min_args_base;
612 #endif // CONFIG_VP9_HIGHBITDEPTH
613  double sum_bitrate = 0.0;
614  double sum_bitrate2 = 0.0;
615  double framerate = 30.0;
616 
617  zero(rc.layer_target_bitrate);
618  memset(&layer_id, 0, sizeof(vpx_svc_layer_id_t));
619  memset(&input_ctx, 0, sizeof(input_ctx));
620  /* Setup default input stream settings */
621  input_ctx.framerate.numerator = 30;
622  input_ctx.framerate.denominator = 1;
623  input_ctx.only_i420 = 1;
624  input_ctx.bit_depth = 0;
625 
626  exec_name = argv[0];
627  // Check usage and arguments.
628  if (argc < min_args) {
629 #if CONFIG_VP9_HIGHBITDEPTH
630  die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
631  "<rate_num> <rate_den> <speed> <frame_drop_threshold> "
632  "<error_resilient> <threads> <mode> "
633  "<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n",
634  argv[0]);
635 #else
636  die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
637  "<rate_num> <rate_den> <speed> <frame_drop_threshold> "
638  "<error_resilient> <threads> <mode> "
639  "<Rate_0> ... <Rate_nlayers-1> \n",
640  argv[0]);
641 #endif // CONFIG_VP9_HIGHBITDEPTH
642  }
643 
644  encoder = get_vpx_encoder_by_name(argv[3]);
645  if (!encoder) die("Unsupported codec.");
646 
647  printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
648 
649  width = (unsigned int)strtoul(argv[4], NULL, 0);
650  height = (unsigned int)strtoul(argv[5], NULL, 0);
651  if (width < 16 || width % 2 || height < 16 || height % 2) {
652  die("Invalid resolution: %d x %d", width, height);
653  }
654 
655  layering_mode = (int)strtol(argv[12], NULL, 0);
656  if (layering_mode < 0 || layering_mode > 13) {
657  die("Invalid layering mode (0..12) %s", argv[12]);
658  }
659 
660  if (argc != min_args + mode_to_num_layers[layering_mode]) {
661  die("Invalid number of arguments");
662  }
663 
664  input_ctx.filename = argv[1];
665  open_input_file(&input_ctx);
666 
667 #if CONFIG_VP9_HIGHBITDEPTH
668  switch (strtol(argv[argc - 1], NULL, 0)) {
669  case 8:
670  bit_depth = VPX_BITS_8;
671  input_bit_depth = 8;
672  break;
673  case 10:
674  bit_depth = VPX_BITS_10;
675  input_bit_depth = 10;
676  break;
677  case 12:
678  bit_depth = VPX_BITS_12;
679  input_bit_depth = 12;
680  break;
681  default: die("Invalid bit depth (8, 10, 12) %s", argv[argc - 1]);
682  }
683 
684  // Y4M reader has its own allocation.
685  if (input_ctx.file_type != FILE_TYPE_Y4M) {
686  if (!vpx_img_alloc(
687  &raw,
689  width, height, 32)) {
690  die("Failed to allocate image", width, height);
691  }
692  }
693 #else
694  // Y4M reader has its own allocation.
695  if (input_ctx.file_type != FILE_TYPE_Y4M) {
696  if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
697  die("Failed to allocate image", width, height);
698  }
699  }
700 #endif // CONFIG_VP9_HIGHBITDEPTH
701 
702  // Populate encoder configuration.
703  res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
704  if (res) {
705  printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
706  return EXIT_FAILURE;
707  }
708 
709  // Update the default configuration with our settings.
710  cfg.g_w = width;
711  cfg.g_h = height;
712 
713 #if CONFIG_VP9_HIGHBITDEPTH
714  if (bit_depth != VPX_BITS_8) {
715  cfg.g_bit_depth = bit_depth;
716  cfg.g_input_bit_depth = input_bit_depth;
717  cfg.g_profile = 2;
718  }
719 #endif // CONFIG_VP9_HIGHBITDEPTH
720 
721  // Timebase format e.g. 30fps: numerator=1, demoninator = 30.
722  cfg.g_timebase.num = (int)strtol(argv[6], NULL, 0);
723  cfg.g_timebase.den = (int)strtol(argv[7], NULL, 0);
724 
725  speed = (int)strtol(argv[8], NULL, 0);
726  if (speed < 0) {
727  die("Invalid speed setting: must be positive");
728  }
729  if (strncmp(encoder->name, "vp9", 3) == 0 && speed > 9) {
730  warn("Mapping speed %d to speed 9.\n", speed);
731  }
732 
733  for (i = min_args_base;
734  (int)i < min_args_base + mode_to_num_layers[layering_mode]; ++i) {
735  rc.layer_target_bitrate[i - 13] = (int)strtol(argv[i], NULL, 0);
736  if (strncmp(encoder->name, "vp8", 3) == 0)
737  cfg.ts_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
738  else if (strncmp(encoder->name, "vp9", 3) == 0)
739  cfg.layer_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
740  }
741 
742  // Real time parameters.
743  cfg.rc_dropframe_thresh = (unsigned int)strtoul(argv[9], NULL, 0);
744  cfg.rc_end_usage = VPX_CBR;
745  cfg.rc_min_quantizer = 2;
746  cfg.rc_max_quantizer = 56;
747  if (strncmp(encoder->name, "vp9", 3) == 0) cfg.rc_max_quantizer = 52;
748  cfg.rc_undershoot_pct = 50;
749  cfg.rc_overshoot_pct = 50;
750  cfg.rc_buf_initial_sz = 600;
751  cfg.rc_buf_optimal_sz = 600;
752  cfg.rc_buf_sz = 1000;
753 
754  // Disable dynamic resizing by default.
755  cfg.rc_resize_allowed = 0;
756 
757  // Use 1 thread as default.
758  cfg.g_threads = (unsigned int)strtoul(argv[11], NULL, 0);
759 
760  error_resilient = (uint32_t)strtoul(argv[10], NULL, 0);
761  if (error_resilient != 0 && error_resilient != 1) {
762  die("Invalid value for error resilient (0, 1): %d.", error_resilient);
763  }
764  // Enable error resilient mode.
765  cfg.g_error_resilient = error_resilient;
766  cfg.g_lag_in_frames = 0;
767  cfg.kf_mode = VPX_KF_AUTO;
768 
769  // Disable automatic keyframe placement.
770  cfg.kf_min_dist = cfg.kf_max_dist = 3000;
771 
773 
774  set_temporal_layer_pattern(layering_mode, &cfg, layer_flags,
775  &flag_periodicity);
776 
777  set_rate_control_metrics(&rc, &cfg);
778 
779  if (input_ctx.file_type == FILE_TYPE_Y4M) {
780  if (input_ctx.width != cfg.g_w || input_ctx.height != cfg.g_h) {
781  die("Incorrect width or height: %d x %d", cfg.g_w, cfg.g_h);
782  }
783  if (input_ctx.framerate.numerator != cfg.g_timebase.den ||
784  input_ctx.framerate.denominator != cfg.g_timebase.num) {
785  die("Incorrect framerate: numerator %d denominator %d",
786  cfg.g_timebase.num, cfg.g_timebase.den);
787  }
788  }
789 
790  framerate = cfg.g_timebase.den / cfg.g_timebase.num;
791  // Open an output file for each stream.
792  for (i = 0; i < cfg.ts_number_layers; ++i) {
793  char file_name[PATH_MAX];
794  VpxVideoInfo info;
795  info.codec_fourcc = encoder->fourcc;
796  info.frame_width = cfg.g_w;
797  info.frame_height = cfg.g_h;
798  info.time_base.numerator = cfg.g_timebase.num;
799  info.time_base.denominator = cfg.g_timebase.den;
800 
801  snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
802  outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
803  if (!outfile[i]) die("Failed to open %s for writing", file_name);
804 
805  assert(outfile[i] != NULL);
806  }
807  // No spatial layers in this encoder.
808  cfg.ss_number_layers = 1;
809 
810 // Initialize codec.
811 #if CONFIG_VP9_HIGHBITDEPTH
812  if (vpx_codec_enc_init(
813  &codec, encoder->codec_interface(), &cfg,
814  bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
815 #else
816  if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
817 #endif // CONFIG_VP9_HIGHBITDEPTH
818  die("Failed to initialize encoder");
819 
820  if (strncmp(encoder->name, "vp8", 3) == 0) {
821  vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
822  vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kVp8DenoiserOff);
825 #if ROI_MAP
826  set_roi_map(encoder->name, &cfg, &roi);
827  if (vpx_codec_control(&codec, VP8E_SET_ROI_MAP, &roi))
828  die_codec(&codec, "Failed to set ROI map");
829 #endif
830 
831  } else if (strncmp(encoder->name, "vp9", 3) == 0) {
832  vpx_svc_extra_cfg_t svc_params;
833  memset(&svc_params, 0, sizeof(svc_params));
834  vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
839  vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kVp9DenoiserOff);
842  vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, get_msb(cfg.g_threads));
843 #if ROI_MAP
844  set_roi_map(encoder->name, &cfg, &roi);
845  if (vpx_codec_control(&codec, VP9E_SET_ROI_MAP, &roi))
846  die_codec(&codec, "Failed to set ROI map");
848 #endif
849  if (cfg.g_threads > 1)
851  else
853  if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1 : 0))
854  die_codec(&codec, "Failed to set SVC");
855  for (i = 0; i < cfg.ts_number_layers; ++i) {
856  svc_params.max_quantizers[i] = cfg.rc_max_quantizer;
857  svc_params.min_quantizers[i] = cfg.rc_min_quantizer;
858  }
859  svc_params.scaling_factor_num[0] = cfg.g_h;
860  svc_params.scaling_factor_den[0] = cfg.g_h;
861  vpx_codec_control(&codec, VP9E_SET_SVC_PARAMETERS, &svc_params);
862  }
863  if (strncmp(encoder->name, "vp8", 3) == 0) {
865  }
867  // This controls the maximum target size of the key frame.
868  // For generating smaller key frames, use a smaller max_intra_size_pct
869  // value, like 100 or 200.
870  {
871  const int max_intra_size_pct = 1000;
873  max_intra_size_pct);
874  }
875 
876  frame_avail = 1;
877  while (frame_avail || got_data) {
878  struct vpx_usec_timer timer;
879  vpx_codec_iter_t iter = NULL;
880  const vpx_codec_cx_pkt_t *pkt;
881  // Update the temporal layer_id. No spatial layers in this test.
882  layer_id.spatial_layer_id = 0;
883  layer_id.temporal_layer_id =
884  cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
885  layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
886  if (strncmp(encoder->name, "vp9", 3) == 0) {
887  vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
888  } else if (strncmp(encoder->name, "vp8", 3) == 0) {
890  layer_id.temporal_layer_id);
891  }
892  flags = layer_flags[frame_cnt % flag_periodicity];
893  if (layering_mode == 0) flags = 0;
894  frame_avail = read_frame(&input_ctx, &raw);
895  if (frame_avail) ++rc.layer_input_frames[layer_id.temporal_layer_id];
896  vpx_usec_timer_start(&timer);
897  if (vpx_codec_encode(&codec, frame_avail ? &raw : NULL, pts, 1, flags,
898  VPX_DL_REALTIME)) {
899  die_codec(&codec, "Failed to encode frame");
900  }
901  vpx_usec_timer_mark(&timer);
902  cx_time += vpx_usec_timer_elapsed(&timer);
903  // Reset KF flag.
904  if (layering_mode != 7) {
905  layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
906  }
907  got_data = 0;
908  while ((pkt = vpx_codec_get_cx_data(&codec, &iter))) {
909  got_data = 1;
910  switch (pkt->kind) {
912  for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
913  i < cfg.ts_number_layers; ++i) {
914  vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
915  pkt->data.frame.sz, pts);
916  ++rc.layer_tot_enc_frames[i];
917  rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
918  // Keep count of rate control stats per layer (for non-key frames).
919  if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
920  !(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
921  rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
922  rc.layer_avg_rate_mismatch[i] +=
923  fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
924  rc.layer_pfb[i];
925  ++rc.layer_enc_frames[i];
926  }
927  }
928  // Update for short-time encoding bitrate states, for moving window
929  // of size rc->window, shifted by rc->window / 2.
930  // Ignore first window segment, due to key frame.
931  if (frame_cnt > rc.window_size) {
932  sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
933  if (frame_cnt % rc.window_size == 0) {
934  rc.window_count += 1;
935  rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
936  rc.variance_st_encoding_bitrate +=
937  (sum_bitrate / rc.window_size) *
938  (sum_bitrate / rc.window_size);
939  sum_bitrate = 0.0;
940  }
941  }
942  // Second shifted window.
943  if (frame_cnt > rc.window_size + rc.window_size / 2) {
944  sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
945  if (frame_cnt > 2 * rc.window_size &&
946  frame_cnt % rc.window_size == 0) {
947  rc.window_count += 1;
948  rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
949  rc.variance_st_encoding_bitrate +=
950  (sum_bitrate2 / rc.window_size) *
951  (sum_bitrate2 / rc.window_size);
952  sum_bitrate2 = 0.0;
953  }
954  }
955  break;
956  default: break;
957  }
958  }
959  ++frame_cnt;
960  pts += frame_duration;
961  }
962  close_input_file(&input_ctx);
963  printout_rate_control_summary(&rc, &cfg, frame_cnt);
964  printf("\n");
965  printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
966  frame_cnt, 1000 * (float)cx_time / (double)(frame_cnt * 1000000),
967  1000000 * (double)frame_cnt / (double)cx_time);
968 
969  if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
970 
971  // Try to rewrite the output file headers with the actual frame count.
972  for (i = 0; i < cfg.ts_number_layers; ++i) vpx_video_writer_close(outfile[i]);
973 
974  if (input_ctx.file_type != FILE_TYPE_Y4M) {
975  vpx_img_free(&raw);
976  }
977 
978 #if ROI_MAP
979  free(roi.roi_map);
980 #endif
981  return EXIT_SUCCESS;
982 }
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:190
enum vpx_bit_depth vpx_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
const char * vpx_codec_iface_name(vpx_codec_iface_t *iface)
Return the name for a given interface.
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition: vpx_codec.h:407
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:93
@ VPX_BITS_8
Definition: vpx_codec.h:221
@ VPX_BITS_12
Definition: vpx_codec.h:223
@ VPX_BITS_10
Definition: vpx_codec.h:222
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition: vpx_encoder.h:830
#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:741
#define VPX_EFLAG_FORCE_KF
Definition: vpx_encoder.h:260
#define VPX_TS_MAX_PERIODICITY
Definition: vpx_encoder.h:37
#define VPX_CODEC_USE_HIGHBITDEPTH
Definition: vpx_encoder.h:90
#define VPX_MAX_LAYERS
Definition: vpx_encoder.h:43
#define VPX_FRAME_IS_KEY
Definition: vpx_encoder.h:116
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int usage)
Get a default configuration.
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
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:147
@ VPX_KF_AUTO
Definition: vpx_encoder.h:248
@ VPX_CBR
Definition: vpx_encoder.h:233
#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:413
@ VP9E_SET_SVC_LAYER_ID
Codec control function to set svc layer for spatial and temporal.
Definition: vp8cx.h:453
@ VP8E_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set Max data rate for Intra frames.
Definition: vp8cx.h:257
@ VP9E_SET_ROI_MAP
Codec control function to pass an ROI map to encoder.
Definition: vp8cx.h:436
@ 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:398
@ VP8E_SET_NOISE_SENSITIVITY
control function to set noise sensitivity
Definition: vp8cx.h:173
@ VP8E_SET_TOKEN_PARTITIONS
Codec control function to set the number of token partitions.
Definition: vp8cx.h:194
@ VP9E_SET_SVC_PARAMETERS
Codec control function to set parameters for SVC.
Definition: vp8cx.h:444
@ VP9E_SET_FRAME_PARALLEL_DECODING
Codec control function to enable frame parallel decoding feature.
Definition: vp8cx.h:385
@ VP8E_SET_GF_CBR_BOOST_PCT
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:601
@ VP9E_SET_TUNE_CONTENT
Codec control function to set content type.
Definition: vp8cx.h:463
@ VP9E_SET_SVC
Codec control function to turn on/off SVC in encoder.
Definition: vp8cx.h:430
@ VP9E_SET_ROW_MT
Codec control function to set row level multi-threading.
Definition: vp8cx.h:570
@ 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:304
@ VP9E_SET_TILE_COLUMNS
Codec control function to set number of tile columns.
Definition: vp8cx.h:351
@ VP8E_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static.
Definition: vp8cx.h:188
@ VP8E_SET_SCREEN_CONTENT_MODE
Codec control function to set encoder screen content mode.
Definition: vp8cx.h:312
@ VP9E_SET_NOISE_SENSITIVITY
Codec control function to set noise sensitivity.
Definition: vp8cx.h:421
@ VP9E_SET_GF_CBR_BOOST_PCT
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:293
@ 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:716
Codec context structure.
Definition: vpx_codec.h:200
Encoder output packet.
Definition: vpx_encoder.h:159
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:160
struct vpx_codec_cx_pkt::@1::@2 frame
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition: vpx_encoder.h:268
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:400
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:693
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:605
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:473
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:644
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:624
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:295
unsigned int layer_target_bitrate[12]
Target bitrate for each spatial/temporal layer.
Definition: vpx_encoder.h:684
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:313
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:596
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition: vpx_encoder.h:676
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:351
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:667
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:516
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:304
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:531
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:391
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:343
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:482
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:372
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:440
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:540
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:321
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:549
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:460
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition: vpx_encoder.h:651
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:329
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:501
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:658
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:614
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:285
Image Descriptor.
Definition: vpx_image.h:72
int den
Definition: vpx_encoder.h:220
int num
Definition: vpx_encoder.h:219
vpx region of interest map
Definition: vp8cx.h:733
int skip[8]
Definition: vp8cx.h:745
unsigned int static_threshold[4]
Definition: vp8cx.h:748
unsigned int rows
Definition: vp8cx.h:739
unsigned int cols
Definition: vp8cx.h:740
int ref_frame[8]
Definition: vp8cx.h:746
int delta_q[8]
Definition: vp8cx.h:742
unsigned char * roi_map
Definition: vp8cx.h:738
int delta_lf[8]
Definition: vp8cx.h:743
vp9 svc layer parameters
Definition: vp8cx.h:810
int temporal_layer_id
Definition: vp8cx.h:813
int temporal_layer_id_per_spatial[5]
Definition: vp8cx.h:814
int spatial_layer_id
Definition: vp8cx.h:811
vp9 svc extra configure parameters
Definition: vpx_encoder.h:701
int min_quantizers[12]
Definition: vpx_encoder.h:703
int scaling_factor_num[12]
Definition: vpx_encoder.h:704
int max_quantizers[12]
Definition: vpx_encoder.h:702
int scaling_factor_den[12]
Definition: vpx_encoder.h:705
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
Describes the encoder algorithm interface to applications.
@ VPX_IMG_FMT_I42016
Definition: vpx_image.h:47
@ VPX_IMG_FMT_I420
Definition: vpx_image.h:42
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.
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.