WebM Codec SDK
vp8_multi_resolution_encoder
1 /*
2  * Copyright (c) 2010 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 /*
12  * This is an example demonstrating multi-resolution encoding in VP8.
13  * High-resolution input video is down-sampled to lower-resolutions. The
14  * encoder then encodes the video and outputs multiple bitstreams with
15  * different resolutions.
16  *
17  * This test also allows for settings temporal layers for each spatial layer.
18  * Different number of temporal layers per spatial stream may be used.
19  * Currently up to 3 temporal layers per spatial stream (encoder) are supported
20  * in this test.
21  */
22 
23 #include "./vpx_config.h"
24 
25 #include <stdio.h>
26 #include <stdlib.h>
27 #include <stdarg.h>
28 #include <string.h>
29 #include <math.h>
30 #include <assert.h>
31 #include <sys/time.h>
32 #include "vpx_ports/vpx_timer.h"
33 #include "vpx/vpx_encoder.h"
34 #include "vpx/vp8cx.h"
35 #include "vpx_ports/mem_ops.h"
36 #include "../tools_common.h"
37 #define interface (vpx_codec_vp8_cx())
38 #define fourcc 0x30385056
39 
40 void usage_exit(void) { exit(EXIT_FAILURE); }
41 
42 /*
43  * The input video frame is downsampled several times to generate a multi-level
44  * hierarchical structure. NUM_ENCODERS is defined as the number of encoding
45  * levels required. For example, if the size of input video is 1280x720,
46  * NUM_ENCODERS is 3, and down-sampling factor is 2, the encoder outputs 3
47  * bitstreams with resolution of 1280x720(level 0), 640x360(level 1), and
48  * 320x180(level 2) respectively.
49  */
50 
51 /* Number of encoders (spatial resolutions) used in this test. */
52 #define NUM_ENCODERS 3
53 
54 /* Maximum number of temporal layers allowed for this test. */
55 #define MAX_NUM_TEMPORAL_LAYERS 3
56 
57 /* This example uses the scaler function in libyuv. */
58 #include "third_party/libyuv/include/libyuv/basic_types.h"
59 #include "third_party/libyuv/include/libyuv/scale.h"
60 #include "third_party/libyuv/include/libyuv/cpu_id.h"
61 
62 int (*read_frame_p)(FILE *f, vpx_image_t *img);
63 
64 static int mulres_read_frame(FILE *f, vpx_image_t *img) {
65  size_t nbytes, to_read;
66  int res = 1;
67 
68  to_read = img->w * img->h * 3 / 2;
69  nbytes = fread(img->planes[0], 1, to_read, f);
70  if (nbytes != to_read) {
71  res = 0;
72  if (nbytes > 0)
73  printf("Warning: Read partial frame. Check your width & height!\n");
74  }
75  return res;
76 }
77 
78 static int mulres_read_frame_by_row(FILE *f, vpx_image_t *img) {
79  size_t nbytes, to_read;
80  int res = 1;
81  int plane;
82 
83  for (plane = 0; plane < 3; plane++) {
84  unsigned char *ptr;
85  int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
86  int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
87  int r;
88 
89  /* Determine the correct plane based on the image format. The for-loop
90  * always counts in Y,U,V order, but this may not match the order of
91  * the data on disk.
92  */
93  switch (plane) {
94  case 1:
96  : VPX_PLANE_U];
97  break;
98  case 2:
100  : VPX_PLANE_V];
101  break;
102  default: ptr = img->planes[plane];
103  }
104 
105  for (r = 0; r < h; r++) {
106  to_read = w;
107 
108  nbytes = fread(ptr, 1, to_read, f);
109  if (nbytes != to_read) {
110  res = 0;
111  if (nbytes > 0)
112  printf("Warning: Read partial frame. Check your width & height!\n");
113  break;
114  }
115 
116  ptr += img->stride[plane];
117  }
118  if (!res) break;
119  }
120 
121  return res;
122 }
123 
124 static void write_ivf_file_header(FILE *outfile, const vpx_codec_enc_cfg_t *cfg,
125  int frame_cnt) {
126  char header[32];
127 
128  if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS) return;
129  header[0] = 'D';
130  header[1] = 'K';
131  header[2] = 'I';
132  header[3] = 'F';
133  mem_put_le16(header + 4, 0); /* version */
134  mem_put_le16(header + 6, 32); /* headersize */
135  mem_put_le32(header + 8, fourcc); /* headersize */
136  mem_put_le16(header + 12, cfg->g_w); /* width */
137  mem_put_le16(header + 14, cfg->g_h); /* height */
138  mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
139  mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
140  mem_put_le32(header + 24, frame_cnt); /* length */
141  mem_put_le32(header + 28, 0); /* unused */
142 
143  (void)fwrite(header, 1, 32, outfile);
144 }
145 
146 static void write_ivf_frame_header(FILE *outfile,
147  const vpx_codec_cx_pkt_t *pkt) {
148  char header[12];
149  vpx_codec_pts_t pts;
150 
151  if (pkt->kind != VPX_CODEC_CX_FRAME_PKT) return;
152 
153  pts = pkt->data.frame.pts;
154  mem_put_le32(header, (int)pkt->data.frame.sz);
155  mem_put_le32(header + 4, pts & 0xFFFFFFFF);
156  mem_put_le32(header + 8, pts >> 32);
157 
158  (void)fwrite(header, 1, 12, outfile);
159 }
160 
161 /* Temporal scaling parameters */
162 /* This sets all the temporal layer parameters given |num_temporal_layers|,
163  * including the target bit allocation across temporal layers. Bit allocation
164  * parameters will be passed in as user parameters in another version.
165  */
166 static void set_temporal_layer_pattern(int num_temporal_layers,
167  vpx_codec_enc_cfg_t *cfg, int bitrate,
168  int *layer_flags) {
169  assert(num_temporal_layers <= MAX_NUM_TEMPORAL_LAYERS);
170  switch (num_temporal_layers) {
171  case 1: {
172  /* 1-layer */
173  cfg->ts_number_layers = 1;
174  cfg->ts_periodicity = 1;
175  cfg->ts_rate_decimator[0] = 1;
176  cfg->ts_layer_id[0] = 0;
177  cfg->ts_target_bitrate[0] = bitrate;
178 
179  // Update L only.
180  layer_flags[0] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
181  break;
182  }
183 
184  case 2: {
185  /* 2-layers, with sync point at first frame of layer 1. */
186  cfg->ts_number_layers = 2;
187  cfg->ts_periodicity = 2;
188  cfg->ts_rate_decimator[0] = 2;
189  cfg->ts_rate_decimator[1] = 1;
190  cfg->ts_layer_id[0] = 0;
191  cfg->ts_layer_id[1] = 1;
192  // Use 60/40 bit allocation as example.
193  cfg->ts_target_bitrate[0] = (int)(0.6f * bitrate);
194  cfg->ts_target_bitrate[1] = bitrate;
195 
196  /* 0=L, 1=GF */
197  // ARF is used as predictor for all frames, and is only updated on
198  // key frame. Sync point every 8 frames.
199 
200  // Layer 0: predict from L and ARF, update L and G.
201  layer_flags[0] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_ARF;
202 
203  // Layer 1: sync point: predict from L and ARF, and update G.
204  layer_flags[1] =
206 
207  // Layer 0, predict from L and ARF, update L.
208  layer_flags[2] =
210 
211  // Layer 1: predict from L, G and ARF, and update G.
212  layer_flags[3] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
214 
215  // Layer 0
216  layer_flags[4] = layer_flags[2];
217 
218  // Layer 1
219  layer_flags[5] = layer_flags[3];
220 
221  // Layer 0
222  layer_flags[6] = layer_flags[4];
223 
224  // Layer 1
225  layer_flags[7] = layer_flags[5];
226  break;
227  }
228 
229  case 3:
230  default: {
231  // 3-layers structure where ARF is used as predictor for all frames,
232  // and is only updated on key frame.
233  // Sync points for layer 1 and 2 every 8 frames.
234  cfg->ts_number_layers = 3;
235  cfg->ts_periodicity = 4;
236  cfg->ts_rate_decimator[0] = 4;
237  cfg->ts_rate_decimator[1] = 2;
238  cfg->ts_rate_decimator[2] = 1;
239  cfg->ts_layer_id[0] = 0;
240  cfg->ts_layer_id[1] = 2;
241  cfg->ts_layer_id[2] = 1;
242  cfg->ts_layer_id[3] = 2;
243  // Use 45/20/35 bit allocation as example.
244  cfg->ts_target_bitrate[0] = (int)(0.45f * bitrate);
245  cfg->ts_target_bitrate[1] = (int)(0.65f * bitrate);
246  cfg->ts_target_bitrate[2] = bitrate;
247 
248  /* 0=L, 1=GF, 2=ARF */
249 
250  // Layer 0: predict from L and ARF; update L and G.
251  layer_flags[0] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF;
252 
253  // Layer 2: sync point: predict from L and ARF; update none.
254  layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
257 
258  // Layer 1: sync point: predict from L and ARF; update G.
259  layer_flags[2] =
261 
262  // Layer 2: predict from L, G, ARF; update none.
263  layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
265 
266  // Layer 0: predict from L and ARF; update L.
267  layer_flags[4] =
269 
270  // Layer 2: predict from L, G, ARF; update none.
271  layer_flags[5] = layer_flags[3];
272 
273  // Layer 1: predict from L, G, ARF; update G.
274  layer_flags[6] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
275 
276  // Layer 2: predict from L, G, ARF; update none.
277  layer_flags[7] = layer_flags[3];
278  break;
279  }
280  }
281 }
282 
283 /* The periodicity of the pattern given the number of temporal layers. */
284 static int periodicity_to_num_layers[MAX_NUM_TEMPORAL_LAYERS] = { 1, 8, 8 };
285 
286 int main(int argc, char **argv) {
287  FILE *infile, *outfile[NUM_ENCODERS];
288  FILE *downsampled_input[NUM_ENCODERS - 1];
289  char filename[50];
290  vpx_codec_ctx_t codec[NUM_ENCODERS];
291  vpx_codec_enc_cfg_t cfg[NUM_ENCODERS];
292  int frame_cnt = 0;
293  vpx_image_t raw[NUM_ENCODERS];
294  vpx_codec_err_t res[NUM_ENCODERS];
295 
296  int i;
297  int width;
298  int height;
299  int length_frame;
300  int frame_avail;
301  int got_data;
302  int flags = 0;
303  int layer_id = 0;
304 
305  int layer_flags[VPX_TS_MAX_PERIODICITY * NUM_ENCODERS] = { 0 };
306  int flag_periodicity;
307 
308  /*Currently, only realtime mode is supported in multi-resolution encoding.*/
309  int arg_deadline = VPX_DL_REALTIME;
310 
311  /* Set show_psnr to 1/0 to show/not show PSNR. Choose show_psnr=0 if you
312  don't need to know PSNR, which will skip PSNR calculation and save
313  encoding time. */
314  int show_psnr = 0;
315  int key_frame_insert = 0;
316  uint64_t psnr_sse_total[NUM_ENCODERS] = { 0 };
317  uint64_t psnr_samples_total[NUM_ENCODERS] = { 0 };
318  double psnr_totals[NUM_ENCODERS][4] = { { 0, 0 } };
319  int psnr_count[NUM_ENCODERS] = { 0 };
320 
321  int64_t cx_time = 0;
322 
323  /* Set the required target bitrates for each resolution level.
324  * If target bitrate for highest-resolution level is set to 0,
325  * (i.e. target_bitrate[0]=0), we skip encoding at that level.
326  */
327  unsigned int target_bitrate[NUM_ENCODERS] = { 1000, 500, 100 };
328 
329  /* Enter the frame rate of the input video */
330  int framerate = 30;
331 
332  /* Set down-sampling factor for each resolution level.
333  dsf[0] controls down sampling from level 0 to level 1;
334  dsf[1] controls down sampling from level 1 to level 2;
335  dsf[2] is not used. */
336  vpx_rational_t dsf[NUM_ENCODERS] = { { 2, 1 }, { 2, 1 }, { 1, 1 } };
337 
338  /* Set the number of temporal layers for each encoder/resolution level,
339  * starting from highest resoln down to lowest resoln. */
340  unsigned int num_temporal_layers[NUM_ENCODERS] = { 3, 3, 3 };
341 
342  if (argc != (7 + 3 * NUM_ENCODERS))
343  die("Usage: %s <width> <height> <frame_rate> <infile> <outfile(s)> "
344  "<rate_encoder(s)> <temporal_layer(s)> <key_frame_insert> <output "
345  "psnr?> \n",
346  argv[0]);
347 
348  printf("Using %s\n", vpx_codec_iface_name(interface));
349 
350  width = (int)strtol(argv[1], NULL, 0);
351  height = (int)strtol(argv[2], NULL, 0);
352  framerate = (int)strtol(argv[3], NULL, 0);
353 
354  if (width < 16 || width % 2 || height < 16 || height % 2)
355  die("Invalid resolution: %dx%d", width, height);
356 
357  /* Open input video file for encoding */
358  if (!(infile = fopen(argv[4], "rb")))
359  die("Failed to open %s for reading", argv[4]);
360 
361  /* Open output file for each encoder to output bitstreams */
362  for (i = 0; i < NUM_ENCODERS; i++) {
363  if (!target_bitrate[i]) {
364  outfile[i] = NULL;
365  continue;
366  }
367 
368  if (!(outfile[i] = fopen(argv[i + 5], "wb")))
369  die("Failed to open %s for writing", argv[i + 4]);
370  }
371 
372  // Bitrates per spatial layer: overwrite default rates above.
373  for (i = 0; i < NUM_ENCODERS; i++) {
374  target_bitrate[i] = (int)strtol(argv[NUM_ENCODERS + 5 + i], NULL, 0);
375  }
376 
377  // Temporal layers per spatial layers: overwrite default settings above.
378  for (i = 0; i < NUM_ENCODERS; i++) {
379  num_temporal_layers[i] =
380  (int)strtol(argv[2 * NUM_ENCODERS + 5 + i], NULL, 0);
381  if (num_temporal_layers[i] < 1 || num_temporal_layers[i] > 3)
382  die("Invalid temporal layers: %d, Must be 1, 2, or 3. \n",
383  num_temporal_layers[i]);
384  }
385 
386  /* Open file to write out each spatially downsampled input stream. */
387  for (i = 0; i < NUM_ENCODERS - 1; i++) {
388  // Highest resoln is encoder 0.
389  if (sprintf(filename, "ds%d.yuv", NUM_ENCODERS - i) < 0) {
390  return EXIT_FAILURE;
391  }
392  downsampled_input[i] = fopen(filename, "wb");
393  }
394 
395  key_frame_insert = (int)strtol(argv[3 * NUM_ENCODERS + 5], NULL, 0);
396 
397  show_psnr = (int)strtol(argv[3 * NUM_ENCODERS + 6], NULL, 0);
398 
399  /* Populate default encoder configuration */
400  for (i = 0; i < NUM_ENCODERS; i++) {
401  res[i] = vpx_codec_enc_config_default(interface, &cfg[i], 0);
402  if (res[i]) {
403  printf("Failed to get config: %s\n", vpx_codec_err_to_string(res[i]));
404  return EXIT_FAILURE;
405  }
406  }
407 
408  /*
409  * Update the default configuration according to needs of the application.
410  */
411  /* Highest-resolution encoder settings */
412  cfg[0].g_w = width;
413  cfg[0].g_h = height;
414  cfg[0].rc_dropframe_thresh = 0;
415  cfg[0].rc_end_usage = VPX_CBR;
416  cfg[0].rc_resize_allowed = 0;
417  cfg[0].rc_min_quantizer = 2;
418  cfg[0].rc_max_quantizer = 56;
419  cfg[0].rc_undershoot_pct = 100;
420  cfg[0].rc_overshoot_pct = 15;
421  cfg[0].rc_buf_initial_sz = 500;
422  cfg[0].rc_buf_optimal_sz = 600;
423  cfg[0].rc_buf_sz = 1000;
424  cfg[0].g_error_resilient = 1; /* Enable error resilient mode */
425  cfg[0].g_lag_in_frames = 0;
426 
427  /* Disable automatic keyframe placement */
428  /* Note: These 3 settings are copied to all levels. But, except the lowest
429  * resolution level, all other levels are set to VPX_KF_DISABLED internally.
430  */
431  cfg[0].kf_mode = VPX_KF_AUTO;
432  cfg[0].kf_min_dist = 3000;
433  cfg[0].kf_max_dist = 3000;
434 
435  cfg[0].rc_target_bitrate = target_bitrate[0]; /* Set target bitrate */
436  cfg[0].g_timebase.num = 1; /* Set fps */
437  cfg[0].g_timebase.den = framerate;
438 
439  /* Other-resolution encoder settings */
440  for (i = 1; i < NUM_ENCODERS; i++) {
441  memcpy(&cfg[i], &cfg[0], sizeof(vpx_codec_enc_cfg_t));
442 
443  cfg[i].rc_target_bitrate = target_bitrate[i];
444 
445  /* Note: Width & height of other-resolution encoders are calculated
446  * from the highest-resolution encoder's size and the corresponding
447  * down_sampling_factor.
448  */
449  {
450  unsigned int iw = cfg[i - 1].g_w * dsf[i - 1].den + dsf[i - 1].num - 1;
451  unsigned int ih = cfg[i - 1].g_h * dsf[i - 1].den + dsf[i - 1].num - 1;
452  cfg[i].g_w = iw / dsf[i - 1].num;
453  cfg[i].g_h = ih / dsf[i - 1].num;
454  }
455 
456  /* Make width & height to be multiplier of 2. */
457  // Should support odd size ???
458  if ((cfg[i].g_w) % 2) cfg[i].g_w++;
459  if ((cfg[i].g_h) % 2) cfg[i].g_h++;
460  }
461 
462  // Set the number of threads per encode/spatial layer.
463  // (1, 1, 1) means no encoder threading.
464  cfg[0].g_threads = 1;
465  cfg[1].g_threads = 1;
466  cfg[2].g_threads = 1;
467 
468  /* Allocate image for each encoder */
469  for (i = 0; i < NUM_ENCODERS; i++)
470  if (!vpx_img_alloc(&raw[i], VPX_IMG_FMT_I420, cfg[i].g_w, cfg[i].g_h, 32))
471  die("Failed to allocate image (%dx%d)", cfg[i].g_w, cfg[i].g_h);
472 
473  if (raw[0].stride[VPX_PLANE_Y] == (int)raw[0].d_w)
474  read_frame_p = mulres_read_frame;
475  else
476  read_frame_p = mulres_read_frame_by_row;
477 
478  for (i = 0; i < NUM_ENCODERS; i++)
479  if (outfile[i]) write_ivf_file_header(outfile[i], &cfg[i], 0);
480 
481  /* Temporal layers settings */
482  for (i = 0; i < NUM_ENCODERS; i++) {
483  set_temporal_layer_pattern(num_temporal_layers[i], &cfg[i],
484  cfg[i].rc_target_bitrate,
485  &layer_flags[i * VPX_TS_MAX_PERIODICITY]);
486  }
487 
488  /* Initialize multi-encoder */
489  if (vpx_codec_enc_init_multi(&codec[0], interface, &cfg[0], NUM_ENCODERS,
490  (show_psnr ? VPX_CODEC_USE_PSNR : 0), &dsf[0]))
491  die_codec(&codec[0], "Failed to initialize encoder");
492 
493  /* The extra encoding configuration parameters can be set as follows. */
494  /* Set encoding speed */
495  for (i = 0; i < NUM_ENCODERS; i++) {
496  int speed = -6;
497  /* Lower speed for the lowest resolution. */
498  if (i == NUM_ENCODERS - 1) speed = -4;
499  if (vpx_codec_control(&codec[i], VP8E_SET_CPUUSED, speed))
500  die_codec(&codec[i], "Failed to set cpu_used");
501  }
502 
503  /* Set static threshold = 1 for all encoders */
504  for (i = 0; i < NUM_ENCODERS; i++) {
505  if (vpx_codec_control(&codec[i], VP8E_SET_STATIC_THRESHOLD, 1))
506  die_codec(&codec[i], "Failed to set static threshold");
507  }
508 
509  /* Set NOISE_SENSITIVITY to do TEMPORAL_DENOISING */
510  /* Enable denoising for the highest-resolution encoder. */
512  die_codec(&codec[0], "Failed to set noise_sensitivity");
514  die_codec(&codec[1], "Failed to set noise_sensitivity");
515  for (i = 2; i < NUM_ENCODERS; i++) {
517  die_codec(&codec[i], "Failed to set noise_sensitivity");
518  }
519 
520  /* Set the number of token partitions */
521  for (i = 0; i < NUM_ENCODERS; i++) {
522  if (vpx_codec_control(&codec[i], VP8E_SET_TOKEN_PARTITIONS, 1))
523  die_codec(&codec[i], "Failed to set static threshold");
524  }
525 
526  /* Set the max intra target bitrate */
527  for (i = 0; i < NUM_ENCODERS; i++) {
528  unsigned int max_intra_size_pct =
529  (int)(((double)cfg[0].rc_buf_optimal_sz * 0.5) * framerate / 10);
531  max_intra_size_pct))
532  die_codec(&codec[i], "Failed to set static threshold");
533  // printf("%d %d \n",i,max_intra_size_pct);
534  }
535 
536  frame_avail = 1;
537  got_data = 0;
538 
539  while (frame_avail || got_data) {
540  struct vpx_usec_timer timer;
541  vpx_codec_iter_t iter[NUM_ENCODERS] = { NULL };
542  const vpx_codec_cx_pkt_t *pkt[NUM_ENCODERS];
543 
544  flags = 0;
545  frame_avail = read_frame_p(infile, &raw[0]);
546 
547  if (frame_avail) {
548  for (i = 1; i < NUM_ENCODERS; i++) {
549  /*Scale the image down a number of times by downsampling factor*/
550  /* FilterMode 1 or 2 give better psnr than FilterMode 0. */
551  I420Scale(
552  raw[i - 1].planes[VPX_PLANE_Y], raw[i - 1].stride[VPX_PLANE_Y],
553  raw[i - 1].planes[VPX_PLANE_U], raw[i - 1].stride[VPX_PLANE_U],
554  raw[i - 1].planes[VPX_PLANE_V], raw[i - 1].stride[VPX_PLANE_V],
555  raw[i - 1].d_w, raw[i - 1].d_h, raw[i].planes[VPX_PLANE_Y],
556  raw[i].stride[VPX_PLANE_Y], raw[i].planes[VPX_PLANE_U],
557  raw[i].stride[VPX_PLANE_U], raw[i].planes[VPX_PLANE_V],
558  raw[i].stride[VPX_PLANE_V], raw[i].d_w, raw[i].d_h, 1);
559  /* Write out down-sampled input. */
560  length_frame = cfg[i].g_w * cfg[i].g_h * 3 / 2;
561  if (fwrite(raw[i].planes[0], 1, length_frame,
562  downsampled_input[NUM_ENCODERS - i - 1]) !=
563  (unsigned int)length_frame) {
564  return EXIT_FAILURE;
565  }
566  }
567  }
568 
569  /* Set the flags (reference and update) for all the encoders.*/
570  for (i = 0; i < NUM_ENCODERS; i++) {
571  layer_id = cfg[i].ts_layer_id[frame_cnt % cfg[i].ts_periodicity];
572  flags = 0;
573  flag_periodicity = periodicity_to_num_layers[num_temporal_layers[i] - 1];
574  flags = layer_flags[i * VPX_TS_MAX_PERIODICITY +
575  frame_cnt % flag_periodicity];
576  // Key frame flag for first frame.
577  if (frame_cnt == 0) {
578  flags |= VPX_EFLAG_FORCE_KF;
579  }
580  if (frame_cnt > 0 && frame_cnt == key_frame_insert) {
581  flags = VPX_EFLAG_FORCE_KF;
582  }
583 
584  vpx_codec_control(&codec[i], VP8E_SET_FRAME_FLAGS, flags);
585  vpx_codec_control(&codec[i], VP8E_SET_TEMPORAL_LAYER_ID, layer_id);
586  }
587 
588  /* Encode each frame at multi-levels */
589  /* Note the flags must be set to 0 in the encode call if they are set
590  for each frame with the vpx_codec_control(), as done above. */
591  vpx_usec_timer_start(&timer);
592  if (vpx_codec_encode(&codec[0], frame_avail ? &raw[0] : NULL, frame_cnt, 1,
593  0, arg_deadline)) {
594  die_codec(&codec[0], "Failed to encode frame");
595  }
596  vpx_usec_timer_mark(&timer);
597  cx_time += vpx_usec_timer_elapsed(&timer);
598 
599  for (i = NUM_ENCODERS - 1; i >= 0; i--) {
600  got_data = 0;
601  while ((pkt[i] = vpx_codec_get_cx_data(&codec[i], &iter[i]))) {
602  got_data = 1;
603  switch (pkt[i]->kind) {
605  write_ivf_frame_header(outfile[i], pkt[i]);
606  (void)fwrite(pkt[i]->data.frame.buf, 1, pkt[i]->data.frame.sz,
607  outfile[i]);
608  break;
609  case VPX_CODEC_PSNR_PKT:
610  if (show_psnr) {
611  int j;
612 
613  psnr_sse_total[i] += pkt[i]->data.psnr.sse[0];
614  psnr_samples_total[i] += pkt[i]->data.psnr.samples[0];
615  for (j = 0; j < 4; j++) {
616  psnr_totals[i][j] += pkt[i]->data.psnr.psnr[j];
617  }
618  psnr_count[i]++;
619  }
620 
621  break;
622  default: break;
623  }
624  fflush(stdout);
625  }
626  }
627  frame_cnt++;
628  }
629  printf("\n");
630  printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
631  frame_cnt, 1000 * (float)cx_time / (double)(frame_cnt * 1000000),
632  1000000 * (double)frame_cnt / (double)cx_time);
633 
634  fclose(infile);
635 
636  printf("Processed %ld frames.\n", (long int)frame_cnt - 1);
637  for (i = 0; i < NUM_ENCODERS; i++) {
638  /* Calculate PSNR and print it out */
639  if ((show_psnr) && (psnr_count[i] > 0)) {
640  int j;
641  double ovpsnr =
642  sse_to_psnr(psnr_samples_total[i], 255.0, psnr_sse_total[i]);
643 
644  fprintf(stderr, "\n ENC%d PSNR (Overall/Avg/Y/U/V)", i);
645 
646  fprintf(stderr, " %.3lf", ovpsnr);
647  for (j = 0; j < 4; j++) {
648  fprintf(stderr, " %.3lf", psnr_totals[i][j] / psnr_count[i]);
649  }
650  }
651 
652  if (vpx_codec_destroy(&codec[i]))
653  die_codec(&codec[i], "Failed to destroy codec");
654 
655  vpx_img_free(&raw[i]);
656 
657  if (!outfile[i]) continue;
658 
659  /* Try to rewrite the file header with the actual frame count */
660  if (!fseek(outfile[i], 0, SEEK_SET))
661  write_ivf_file_header(outfile[i], &cfg[i], frame_cnt - 1);
662  fclose(outfile[i]);
663  }
664 
665  return EXIT_SUCCESS;
666 }
VPX_IMG_FMT_YV12
@ VPX_IMG_FMT_YV12
Definition: vpx_image.h:40
vpx_codec_enc_init_multi
#define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf)
Convenience macro for vpx_codec_enc_init_multi_ver()
Definition: vpx_encoder.h:920
vpx_codec_enc_cfg::ts_rate_decimator
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:660
vpx_image::d_h
unsigned int d_h
Definition: vpx_image.h:84
vpx_codec_ctx
Codec context structure.
Definition: vpx_codec.h:200
vpx_codec_err_to_string
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
vpx_codec_cx_pkt::kind
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:162
vpx_img_free
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.
vpx_codec_cx_pkt::data
union vpx_codec_cx_pkt::@1 data
vpx_codec_enc_cfg
Encoder configuration structure.
Definition: vpx_encoder.h:270
vpx_codec_enc_cfg::kf_max_dist
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:616
VPX_PLANE_U
#define VPX_PLANE_U
Definition: vpx_image.h:97
vpx_codec_iface_name
const char * vpx_codec_iface_name(vpx_codec_iface_t *iface)
Return the name for a given interface.
VPX_CODEC_PSNR_PKT
@ VPX_CODEC_PSNR_PKT
Definition: vpx_encoder.h:152
VP8E_SET_NOISE_SENSITIVITY
@ VP8E_SET_NOISE_SENSITIVITY
control function to set noise sensitivity
Definition: vp8cx.h:190
vpx_codec_enc_cfg::rc_buf_sz
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:533
vpx_img_alloc
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_rational::num
int num
Definition: vpx_encoder.h:221
VPX_DL_REALTIME
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition: vpx_encoder.h:978
VP8_EFLAG_NO_UPD_ENTROPY
#define VP8_EFLAG_NO_UPD_ENTROPY
Disable entropy update.
Definition: vp8cx.h:133
vpx_codec_get_cx_data
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_cx_pkt
Encoder output packet.
Definition: vpx_encoder.h:161
vpx_codec_iter_t
const typedef void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:190
vpx_codec_enc_cfg::g_w
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:306
VPX_CODEC_CX_FRAME_PKT
@ VPX_CODEC_CX_FRAME_PKT
Definition: vpx_encoder.h:149
vpx_image::d_w
unsigned int d_w
Definition: vpx_image.h:83
vpx_image::planes
unsigned char * planes[4]
Definition: vpx_image.h:100
vpx_image::fmt
vpx_img_fmt_t fmt
Definition: vpx_image.h:73
VP8_EFLAG_NO_UPD_LAST
#define VP8_EFLAG_NO_UPD_LAST
Don't update the last frame.
Definition: vp8cx.h:98
vpx_codec_enc_cfg::rc_overshoot_pct
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:518
vpx_codec_pts_t
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:108
VPX_RC_LAST_PASS
@ VPX_RC_LAST_PASS
Definition: vpx_encoder.h:229
VP8E_SET_STATIC_THRESHOLD
@ VP8E_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static.
Definition: vp8cx.h:205
vpx_codec_enc_cfg::g_timebase
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:345
VPX_IMG_FMT_I420
@ VPX_IMG_FMT_I420
Definition: vpx_image.h:42
vpx_codec_cx_pkt::psnr
double psnr[4]
Definition: vpx_encoder.h:189
VPX_TS_MAX_PERIODICITY
#define VPX_TS_MAX_PERIODICITY
Definition: vpx_encoder.h:38
vpx_codec_enc_cfg::rc_dropframe_thresh
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:393
vpx_codec_enc_cfg::g_error_resilient
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:353
vpx_image
Image Descriptor.
Definition: vpx_image.h:72
vpx_codec_enc_cfg::g_pass
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: vpx_encoder.h:360
vpx_codec_enc_cfg::g_h
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:315
vpx_image::stride
int stride[4]
Definition: vpx_image.h:101
vpx_codec_enc_cfg::rc_max_quantizer
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:484
vpx_codec_enc_cfg::rc_end_usage
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:442
vpx_codec_enc_cfg::g_lag_in_frames
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:374
VP8E_SET_CPUUSED
@ VP8E_SET_CPUUSED
Codec control function to set encoder internal speed settings.
Definition: vp8cx.h:172
vpx_codec_err_t
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:93
vpx_codec_enc_cfg::rc_resize_allowed
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:402
VPX_EFLAG_FORCE_KF
#define VPX_EFLAG_FORCE_KF
Definition: vpx_encoder.h:262
VPX_PLANE_V
#define VPX_PLANE_V
Definition: vpx_image.h:98
vpx_codec_enc_cfg::ts_target_bitrate
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition: vpx_encoder.h:653
VP8E_SET_FRAME_FLAGS
@ VP8E_SET_FRAME_FLAGS
Codec control function to set reference and update frame flags.
Definition: vp8cx.h:280
vp8cx.h
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
vpx_codec_enc_cfg::kf_min_dist
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:607
vpx_codec_enc_cfg::rc_min_quantizer
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:475
VPX_PLANE_Y
#define VPX_PLANE_Y
Definition: vpx_image.h:96
vpx_image::h
unsigned int h
Definition: vpx_image.h:79
vpx_codec_enc_cfg::rc_buf_initial_sz
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:542
vpx_codec_enc_cfg::rc_buf_optimal_sz
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:551
VP8E_SET_MAX_INTRA_BITRATE_PCT
@ VP8E_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set Max data rate for Intra frames.
Definition: vp8cx.h:274
VPX_KF_AUTO
@ VPX_KF_AUTO
Definition: vpx_encoder.h:250
vpx_codec_enc_cfg::ts_layer_id
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition: vpx_encoder.h:678
vp9_ref_frame::img
vpx_image_t img
Definition: vp8.h:112
VP8E_SET_TOKEN_PARTITIONS
@ VP8E_SET_TOKEN_PARTITIONS
Codec control function to set the number of token partitions.
Definition: vp8cx.h:211
vpx_codec_enc_cfg::rc_target_bitrate
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:462
vpx_codec_enc_cfg::g_threads
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:287
VP8_EFLAG_NO_REF_GF
#define VP8_EFLAG_NO_REF_GF
Don't reference the golden frame.
Definition: vp8cx.h:83
vpx_codec_control
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition: vpx_codec.h:407
VP8_EFLAG_NO_UPD_ARF
#define VP8_EFLAG_NO_UPD_ARF
Don't update the alternate reference frame.
Definition: vp8cx.h:112
vpx_codec_encode
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_pkt::frame
struct vpx_codec_cx_pkt::@1::@2 frame
vpx_rational::den
int den
Definition: vpx_encoder.h:222
vpx_codec_enc_config_default
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.
vpx_codec_destroy
vpx_codec_err_t vpx_codec_destroy(vpx_codec_ctx_t *ctx)
Destroy a codec instance.
VP8_EFLAG_NO_UPD_GF
#define VP8_EFLAG_NO_UPD_GF
Don't update the golden frame.
Definition: vp8cx.h:105
VPX_CODEC_USE_PSNR
#define VPX_CODEC_USE_PSNR
Initialization-time Feature Enabling.
Definition: vpx_encoder.h:89
vpx_encoder.h
Describes the encoder algorithm interface to applications.
vpx_rational
Rational Number.
Definition: vpx_encoder.h:220
VPX_RC_ONE_PASS
@ VPX_RC_ONE_PASS
Definition: vpx_encoder.h:227
VPX_CBR
@ VPX_CBR
Definition: vpx_encoder.h:235
vpx_codec_enc_cfg::ts_number_layers
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:646
vpx_codec_enc_cfg::rc_undershoot_pct
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:503
vpx_image::w
unsigned int w
Definition: vpx_image.h:78
VP8E_SET_TEMPORAL_LAYER_ID
@ VP8E_SET_TEMPORAL_LAYER_ID
Codec control function to set the temporal layer id.
Definition: vp8cx.h:321
vpx_codec_enc_cfg::kf_mode
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:598
vpx_codec_enc_cfg::ts_periodicity
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:669