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/*
12 * This is an example demonstrating how to implement a multi-layer
13 * VP9 encoding scheme based on spatial scalability for video applications
14 * that benefit from a scalable bitstream.
15 */
16
17#include <stdarg.h>
18#include <stdlib.h>
19#include <string.h>
20#include <time.h>
21#include "./args.h"
22#include "vpx/svc_context.h"
23#include "vpx/vp8cx.h"
24#include "vpx/vpx_encoder.h"
25
26#define VP90_FOURCC 0x30395056
27
28static const struct arg_enum_list encoding_mode_enum[] = {
29  {"i", INTER_LAYER_PREDICTION_I},
30  {"alt-ip", ALT_INTER_LAYER_PREDICTION_IP},
31  {"ip", INTER_LAYER_PREDICTION_IP},
32  {"gf", USE_GOLDEN_FRAME},
33  {NULL, 0}
34};
35
36static const arg_def_t encoding_mode_arg = ARG_DEF_ENUM(
37    "m", "encoding-mode", 1, "Encoding mode algorithm", encoding_mode_enum);
38static const arg_def_t skip_frames_arg =
39    ARG_DEF("s", "skip-frames", 1, "input frames to skip");
40static const arg_def_t frames_arg =
41    ARG_DEF("f", "frames", 1, "number of frames to encode");
42static const arg_def_t width_arg = ARG_DEF("w", "width", 1, "source width");
43static const arg_def_t height_arg = ARG_DEF("h", "height", 1, "source height");
44static const arg_def_t timebase_arg =
45    ARG_DEF("t", "timebase", 1, "timebase (num/den)");
46static const arg_def_t bitrate_arg = ARG_DEF(
47    "b", "target-bitrate", 1, "encoding bitrate, in kilobits per second");
48static const arg_def_t layers_arg =
49    ARG_DEF("l", "layers", 1, "number of SVC layers");
50static const arg_def_t kf_dist_arg =
51    ARG_DEF("k", "kf-dist", 1, "number of frames between keyframes");
52static const arg_def_t scale_factors_arg =
53    ARG_DEF("r", "scale-factors", 1, "scale factors (lowest to highest layer)");
54static const arg_def_t quantizers_arg =
55    ARG_DEF("q", "quantizers", 1, "quantizers (lowest to highest layer)");
56static const arg_def_t dummy_frame_arg =
57    ARG_DEF("z", "dummy-frame", 1, "make first frame blank and full size");
58
59static const arg_def_t *svc_args[] = {
60  &encoding_mode_arg, &frames_arg,        &width_arg,       &height_arg,
61  &timebase_arg,      &bitrate_arg,       &skip_frames_arg, &layers_arg,
62  &kf_dist_arg,       &scale_factors_arg, &quantizers_arg,  &dummy_frame_arg,
63  NULL
64};
65
66static const SVC_ENCODING_MODE default_encoding_mode =
67    INTER_LAYER_PREDICTION_IP;
68static const uint32_t default_frames_to_skip = 0;
69static const uint32_t default_frames_to_code = 60 * 60;
70static const uint32_t default_width = 1920;
71static const uint32_t default_height = 1080;
72static const uint32_t default_timebase_num = 1;
73static const uint32_t default_timebase_den = 60;
74static const uint32_t default_bitrate = 1000;
75static const uint32_t default_spatial_layers = 5;
76static const uint32_t default_kf_dist = 100;
77static const int default_use_dummy_frame = 1;
78
79typedef struct {
80  char *input_filename;
81  char *output_filename;
82  uint32_t frames_to_code;
83  uint32_t frames_to_skip;
84} AppInput;
85
86static void mem_put_le16(char *mem, uint32_t val) {
87  mem[0] = val;
88  mem[1] = val >> 8;
89}
90
91static void mem_put_le32(char *mem, uint32_t val) {
92  mem[0] = val;
93  mem[1] = val >> 8;
94  mem[2] = val >> 16;
95  mem[3] = val >> 24;
96}
97
98static void usage(const char *exec_name) {
99  fprintf(stderr, "Usage: %s <options> input_filename output_filename\n",
100          exec_name);
101  fprintf(stderr, "Options:\n");
102  arg_show_usage(stderr, svc_args);
103  exit(EXIT_FAILURE);
104}
105
106void die(const char *fmt, ...) {
107  va_list ap;
108
109  va_start(ap, fmt);
110  vfprintf(stderr, fmt, ap);
111  if (fmt[strlen(fmt) - 1] != '\n') printf("\n");
112  exit(EXIT_FAILURE);
113}
114
115static void die_codec(vpx_codec_ctx_t *ctx, const char *s) {
116  const char *detail = vpx_codec_error_detail(ctx);
117
118  printf("%s: %s\n", s, vpx_codec_error(ctx));
119  if (detail) printf("    %s\n", detail);
120  exit(EXIT_FAILURE);
121}
122
123static int read_frame(FILE *f, vpx_image_t *img) {
124  size_t nbytes;
125  int res = 1;
126  int plane;
127
128  for (plane = 0; plane < 3; ++plane) {
129    uint8_t *ptr;
130    const int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
131    const int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
132    int r;
133
134    switch (plane) {
135      case 1:
136        ptr = img->planes[VPX_PLANE_U];
137        break;
138      case 2:
139        ptr = img->planes[VPX_PLANE_V];
140        break;
141      default:
142        ptr = img->planes[plane];
143    }
144    for (r = 0; r < h; ++r) {
145      const int to_read = w;
146
147      nbytes = fread(ptr, 1, to_read, f);
148      if (nbytes != to_read) {
149        res = 0;
150        if (nbytes > 0)
151          printf("Warning: Read partial frame. Check your width & height!\n");
152        break;
153      }
154      ptr += img->stride[plane];
155    }
156    if (!res) break;
157  }
158  return res;
159}
160
161static int create_dummy_frame(vpx_image_t *img) {
162  const size_t buf_size = img->w * img->h * 3 / 2;
163  memset(img->planes[0], 129, buf_size);
164  return 1;
165}
166
167static void write_ivf_file_header(FILE *outfile,
168                                  uint32_t width, uint32_t height,
169                                  int timebase_num, int timebase_den,
170                                  int frame_cnt) {
171  char header[32];
172
173  header[0] = 'D';
174  header[1] = 'K';
175  header[2] = 'I';
176  header[3] = 'F';
177  mem_put_le16(header + 4, 0);             /* version */
178  mem_put_le16(header + 6, 32);            /* headersize */
179  mem_put_le32(header + 8, VP90_FOURCC);   /* fourcc */
180  mem_put_le16(header + 12, width);        /* width */
181  mem_put_le16(header + 14, height);       /* height */
182  mem_put_le32(header + 16, timebase_den); /* rate */
183  mem_put_le32(header + 20, timebase_num); /* scale */
184  mem_put_le32(header + 24, frame_cnt);    /* length */
185  mem_put_le32(header + 28, 0);            /* unused */
186
187  (void)fwrite(header, 1, 32, outfile);
188}
189
190static void write_ivf_frame_header(FILE *outfile, vpx_codec_pts_t pts,
191                                   size_t sz) {
192  char header[12];
193  mem_put_le32(header, (uint32_t)sz);
194  mem_put_le32(header + 4, pts & 0xFFFFFFFF);
195  mem_put_le32(header + 8, pts >> 32);
196
197  (void)fwrite(header, 1, 12, outfile);
198}
199
200static void parse_command_line(int argc, const char **argv_,
201                               AppInput *app_input, SvcContext *svc_ctx,
202                               vpx_codec_enc_cfg_t *enc_cfg) {
203  struct arg arg;
204  char **argv, **argi, **argj;
205  vpx_codec_err_t res;
206
207  // initialize SvcContext with parameters that will be passed to vpx_svc_init
208  svc_ctx->log_level = SVC_LOG_DEBUG;
209  svc_ctx->spatial_layers = default_spatial_layers;
210  svc_ctx->encoding_mode = default_encoding_mode;
211  // when using a dummy frame, that frame is only encoded to be full size
212  svc_ctx->first_frame_full_size = default_use_dummy_frame;
213
214  // start with default encoder configuration
215  res = vpx_codec_enc_config_default(vpx_codec_vp9_cx(), enc_cfg, 0);
216  if (res) {
217    die("Failed to get config: %s\n", vpx_codec_err_to_string(res));
218  }
219  // update enc_cfg with app default values
220  enc_cfg->g_w = default_width;
221  enc_cfg->g_h = default_height;
222  enc_cfg->g_timebase.num = default_timebase_num;
223  enc_cfg->g_timebase.den = default_timebase_den;
224  enc_cfg->rc_target_bitrate = default_bitrate;
225  enc_cfg->kf_min_dist = default_kf_dist;
226  enc_cfg->kf_max_dist = default_kf_dist;
227
228  // initialize AppInput with default values
229  app_input->frames_to_code = default_frames_to_code;
230  app_input->frames_to_skip = default_frames_to_skip;
231
232  // process command line options
233  argv = argv_dup(argc - 1, argv_ + 1);
234  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
235    arg.argv_step = 1;
236
237    if (arg_match(&arg, &encoding_mode_arg, argi)) {
238      svc_ctx->encoding_mode = arg_parse_enum_or_int(&arg);
239    } else if (arg_match(&arg, &frames_arg, argi)) {
240      app_input->frames_to_code = arg_parse_uint(&arg);
241    } else if (arg_match(&arg, &width_arg, argi)) {
242      enc_cfg->g_w = arg_parse_uint(&arg);
243    } else if (arg_match(&arg, &height_arg, argi)) {
244      enc_cfg->g_h = arg_parse_uint(&arg);
245    } else if (arg_match(&arg, &height_arg, argi)) {
246      enc_cfg->g_h = arg_parse_uint(&arg);
247    } else if (arg_match(&arg, &timebase_arg, argi)) {
248      enc_cfg->g_timebase = arg_parse_rational(&arg);
249    } else if (arg_match(&arg, &bitrate_arg, argi)) {
250      enc_cfg->rc_target_bitrate = arg_parse_uint(&arg);
251    } else if (arg_match(&arg, &skip_frames_arg, argi)) {
252      app_input->frames_to_skip = arg_parse_uint(&arg);
253    } else if (arg_match(&arg, &layers_arg, argi)) {
254      svc_ctx->spatial_layers = arg_parse_uint(&arg);
255    } else if (arg_match(&arg, &kf_dist_arg, argi)) {
256      enc_cfg->kf_min_dist = arg_parse_uint(&arg);
257      enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
258    } else if (arg_match(&arg, &scale_factors_arg, argi)) {
259      vpx_svc_set_scale_factors(svc_ctx, arg.val);
260    } else if (arg_match(&arg, &quantizers_arg, argi)) {
261      vpx_svc_set_quantizers(svc_ctx, arg.val);
262    } else if (arg_match(&arg, &dummy_frame_arg, argi)) {
263      svc_ctx->first_frame_full_size = arg_parse_int(&arg);
264    } else {
265      ++argj;
266    }
267  }
268
269  // Check for unrecognized options
270  for (argi = argv; *argi; ++argi)
271    if (argi[0][0] == '-' && strlen(argi[0]) > 1)
272      die("Error: Unrecognized option %s\n", *argi);
273
274  if (argv[0] == NULL || argv[1] == 0) {
275    usage(argv_[0]);
276  }
277  app_input->input_filename = argv[0];
278  app_input->output_filename = argv[1];
279  free(argv);
280
281  if (enc_cfg->g_w < 16 || enc_cfg->g_w % 2 || enc_cfg->g_h < 16 ||
282      enc_cfg->g_h % 2)
283    die("Invalid resolution: %d x %d\n", enc_cfg->g_w, enc_cfg->g_h);
284
285  printf(
286      "Codec %s\nframes: %d, skip: %d\n"
287      "mode: %d, layers: %d\n"
288      "width %d, height: %d,\n"
289      "num: %d, den: %d, bitrate: %d,\n"
290      "gop size: %d, use_dummy_frame: %d\n",
291      vpx_codec_iface_name(vpx_codec_vp9_cx()), app_input->frames_to_code,
292      app_input->frames_to_skip, svc_ctx->encoding_mode,
293      svc_ctx->spatial_layers, enc_cfg->g_w, enc_cfg->g_h,
294      enc_cfg->g_timebase.num, enc_cfg->g_timebase.den,
295      enc_cfg->rc_target_bitrate, enc_cfg->kf_max_dist,
296      svc_ctx->first_frame_full_size);
297}
298
299int main(int argc, const char **argv) {
300  AppInput app_input = {0};
301  FILE *infile, *outfile;
302  vpx_codec_ctx_t codec;
303  vpx_codec_enc_cfg_t enc_cfg;
304  SvcContext svc_ctx;
305  uint32_t i;
306  uint32_t frame_cnt = 0;
307  vpx_image_t raw;
308  vpx_codec_err_t res;
309  int pts = 0;            /* PTS starts at 0 */
310  int frame_duration = 1; /* 1 timebase tick per frame */
311
312  memset(&svc_ctx, 0, sizeof(svc_ctx));
313  svc_ctx.log_print = 1;
314  parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
315
316  // Allocate image buffer
317  if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32))
318    die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
319
320  if (!(infile = fopen(app_input.input_filename, "rb")))
321    die("Failed to open %s for reading\n", app_input.input_filename);
322
323  if (!(outfile = fopen(app_input.output_filename, "wb")))
324    die("Failed to open %s for writing\n", app_input.output_filename);
325
326  // Initialize codec
327  if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) !=
328      VPX_CODEC_OK)
329    die("Failed to initialize encoder\n");
330
331  write_ivf_file_header(outfile, enc_cfg.g_w, enc_cfg.g_h,
332                        enc_cfg.g_timebase.num, enc_cfg.g_timebase.den, 0);
333
334  // skip initial frames
335  for (i = 0; i < app_input.frames_to_skip; ++i) {
336    read_frame(infile, &raw);
337  }
338
339  // Encode frames
340  while (frame_cnt <= app_input.frames_to_code) {
341    if (frame_cnt == 0 && svc_ctx.first_frame_full_size) {
342      create_dummy_frame(&raw);
343    } else {
344      if (!read_frame(infile, &raw)) break;
345    }
346    res = vpx_svc_encode(&svc_ctx, &codec, &raw, pts, frame_duration,
347                         VPX_DL_REALTIME);
348    printf("%s", vpx_svc_get_message(&svc_ctx));
349    if (res != VPX_CODEC_OK) {
350      die_codec(&codec, "Failed to encode frame");
351    }
352    if (vpx_svc_get_frame_size(&svc_ctx) > 0) {
353      write_ivf_frame_header(outfile, pts, vpx_svc_get_frame_size(&svc_ctx));
354      (void)fwrite(vpx_svc_get_buffer(&svc_ctx), 1,
355                   vpx_svc_get_frame_size(&svc_ctx), outfile);
356    }
357    ++frame_cnt;
358    pts += frame_duration;
359  }
360
361  printf("Processed %d frames\n", frame_cnt - svc_ctx.first_frame_full_size);
362
363  fclose(infile);
364  if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
365
366  // rewrite the output file headers with the actual frame count
367  if (!fseek(outfile, 0, SEEK_SET)) {
368    write_ivf_file_header(outfile, enc_cfg.g_w, enc_cfg.g_h,
369                          enc_cfg.g_timebase.num, enc_cfg.g_timebase.den,
370                          frame_cnt);
371  }
372  fclose(outfile);
373  vpx_img_free(&raw);
374
375  // display average size, psnr
376  printf("%s", vpx_svc_dump_statistics(&svc_ctx));
377
378  vpx_svc_release(&svc_ctx);
379
380  return EXIT_SUCCESS;
381}
382