1233d2500723e5594f3e7c70896ffeeef32b9c950ywan/*
2233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3233d2500723e5594f3e7c70896ffeeef32b9c950ywan *
4233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  Use of this source code is governed by a BSD-style license
5233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  that can be found in the LICENSE file in the root of the source
6233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  tree. An additional intellectual property rights grant can be found
7233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  in the file PATENTS.  All contributing project authors may
8233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  be found in the AUTHORS file in the root of the source tree.
9233d2500723e5594f3e7c70896ffeeef32b9c950ywan */
10233d2500723e5594f3e7c70896ffeeef32b9c950ywan
11233d2500723e5594f3e7c70896ffeeef32b9c950ywan
12233d2500723e5594f3e7c70896ffeeef32b9c950ywan// VP8 Set Reference Frame
13233d2500723e5594f3e7c70896ffeeef32b9c950ywan// =======================
14233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
15233d2500723e5594f3e7c70896ffeeef32b9c950ywan// This is an example demonstrating how to overwrite the VP8 encoder's
16233d2500723e5594f3e7c70896ffeeef32b9c950ywan// internal reference frame. In the sample we set the last frame to the
17233d2500723e5594f3e7c70896ffeeef32b9c950ywan// current frame. If this is done at a cut scene it will avoid a keyframe.
18233d2500723e5594f3e7c70896ffeeef32b9c950ywan// This technique could be used to bounce between two cameras.
19233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
20233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Note that the decoder would also have to set the reference frame to the
21233d2500723e5594f3e7c70896ffeeef32b9c950ywan// same value on the same frame, or the video will become corrupt.
22233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
23233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Usage
24233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -----
25233d2500723e5594f3e7c70896ffeeef32b9c950ywan// This example adds a single argument to the `simple_encoder` example,
26233d2500723e5594f3e7c70896ffeeef32b9c950ywan// which specifies the frame number to update the reference frame on.
27233d2500723e5594f3e7c70896ffeeef32b9c950ywan// The parameter is parsed as follows:
28233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
29233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
30233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Extra Variables
31233d2500723e5594f3e7c70896ffeeef32b9c950ywan// ---------------
32233d2500723e5594f3e7c70896ffeeef32b9c950ywan// This example maintains the frame number passed on the command line
33233d2500723e5594f3e7c70896ffeeef32b9c950ywan// in the `update_frame_num` variable.
34233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
35233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
36233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Configuration
37233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -------------
38233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
39233d2500723e5594f3e7c70896ffeeef32b9c950ywan// The reference frame is updated on the frame specified on the command
40233d2500723e5594f3e7c70896ffeeef32b9c950ywan// line.
41233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
42233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Observing The Effects
43233d2500723e5594f3e7c70896ffeeef32b9c950ywan// ---------------------
44233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Use the `simple_encoder` example to encode a sample with a cut scene.
45233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Determine the frame number of the cut scene by looking for a generated
46233d2500723e5594f3e7c70896ffeeef32b9c950ywan// key-frame (indicated by a 'K'). Supply that frame number as an argument
47233d2500723e5594f3e7c70896ffeeef32b9c950ywan// to this example, and observe that no key-frame is generated.
48233d2500723e5594f3e7c70896ffeeef32b9c950ywan
49233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <stdio.h>
50233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <stdlib.h>
51233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <string.h>
52233d2500723e5594f3e7c70896ffeeef32b9c950ywan
53233d2500723e5594f3e7c70896ffeeef32b9c950ywan#define VPX_CODEC_DISABLE_COMPAT 1
54233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "vpx/vp8cx.h"
55233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "vpx/vpx_encoder.h"
56233d2500723e5594f3e7c70896ffeeef32b9c950ywan
57233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "./tools_common.h"
58233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "./video_writer.h"
59233d2500723e5594f3e7c70896ffeeef32b9c950ywan
60233d2500723e5594f3e7c70896ffeeef32b9c950ywanstatic const char *exec_name;
61233d2500723e5594f3e7c70896ffeeef32b9c950ywan
62233d2500723e5594f3e7c70896ffeeef32b9c950ywanvoid usage_exit() {
63233d2500723e5594f3e7c70896ffeeef32b9c950ywan  fprintf(stderr, "Usage: %s <width> <height> <infile> <outfile> <frame>\n",
64233d2500723e5594f3e7c70896ffeeef32b9c950ywan          exec_name);
65233d2500723e5594f3e7c70896ffeeef32b9c950ywan  exit(EXIT_FAILURE);
66233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
67233d2500723e5594f3e7c70896ffeeef32b9c950ywan
68233d2500723e5594f3e7c70896ffeeef32b9c950ywanstatic void encode_frame(vpx_codec_ctx_t *codec,
69233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         vpx_image_t *img,
70233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         int frame_index,
71233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         VpxVideoWriter *writer) {
72233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_iter_t iter = NULL;
73233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_cx_pkt_t *pkt = NULL;
74233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_err_t res = vpx_codec_encode(codec, img, frame_index, 1, 0,
75233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                               VPX_DL_GOOD_QUALITY);
76233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (res != VPX_CODEC_OK)
77233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(codec, "Failed to encode frame");
78233d2500723e5594f3e7c70896ffeeef32b9c950ywan
79233d2500723e5594f3e7c70896ffeeef32b9c950ywan  while ((pkt = vpx_codec_get_cx_data(codec, &iter)) != NULL) {
80233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) {
81233d2500723e5594f3e7c70896ffeeef32b9c950ywan      const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0;
82233d2500723e5594f3e7c70896ffeeef32b9c950ywan      if (!vpx_video_writer_write_frame(writer,
83233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                        pkt->data.frame.buf,
84233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                        pkt->data.frame.sz,
85233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                        pkt->data.frame.pts)) {
86233d2500723e5594f3e7c70896ffeeef32b9c950ywan        die_codec(codec, "Failed to write compressed frame");
87233d2500723e5594f3e7c70896ffeeef32b9c950ywan      }
88233d2500723e5594f3e7c70896ffeeef32b9c950ywan
89233d2500723e5594f3e7c70896ffeeef32b9c950ywan      printf(keyframe ? "K" : ".");
90233d2500723e5594f3e7c70896ffeeef32b9c950ywan      fflush(stdout);
91233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
92233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
93233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
94233d2500723e5594f3e7c70896ffeeef32b9c950ywan
95233d2500723e5594f3e7c70896ffeeef32b9c950ywanint main(int argc, char **argv) {
96233d2500723e5594f3e7c70896ffeeef32b9c950ywan  FILE *infile = NULL;
97233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_ctx_t codec = {0};
98233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_enc_cfg_t cfg = {0};
99233d2500723e5594f3e7c70896ffeeef32b9c950ywan  int frame_count = 0;
100233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_image_t raw;
101233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_err_t res;
102233d2500723e5594f3e7c70896ffeeef32b9c950ywan  VpxVideoInfo info = {0};
103233d2500723e5594f3e7c70896ffeeef32b9c950ywan  VpxVideoWriter *writer = NULL;
104233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const VpxInterface *encoder = NULL;
105233d2500723e5594f3e7c70896ffeeef32b9c950ywan  int update_frame_num = 0;
106233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const int fps = 30;        // TODO(dkovalev) add command line argument
107233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const int bitrate = 200;   // kbit/s TODO(dkovalev) add command line argument
108233d2500723e5594f3e7c70896ffeeef32b9c950ywan
109233d2500723e5594f3e7c70896ffeeef32b9c950ywan  exec_name = argv[0];
110233d2500723e5594f3e7c70896ffeeef32b9c950ywan
111233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (argc != 6)
112233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Invalid number of arguments");
113233d2500723e5594f3e7c70896ffeeef32b9c950ywan
114233d2500723e5594f3e7c70896ffeeef32b9c950ywan  // TODO(dkovalev): add vp9 support and rename the file accordingly
115233d2500723e5594f3e7c70896ffeeef32b9c950ywan  encoder = get_vpx_encoder_by_name("vp8");
116233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!encoder)
117233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Unsupported codec.");
118233d2500723e5594f3e7c70896ffeeef32b9c950ywan
119233d2500723e5594f3e7c70896ffeeef32b9c950ywan  update_frame_num = atoi(argv[5]);
120233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!update_frame_num)
121233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Couldn't parse frame number '%s'\n", argv[5]);
122233d2500723e5594f3e7c70896ffeeef32b9c950ywan
123233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.codec_fourcc = encoder->fourcc;
124233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.frame_width = strtol(argv[1], NULL, 0);
125233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.frame_height = strtol(argv[2], NULL, 0);
126233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.time_base.numerator = 1;
127233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.time_base.denominator = fps;
128233d2500723e5594f3e7c70896ffeeef32b9c950ywan
129233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (info.frame_width <= 0 ||
130233d2500723e5594f3e7c70896ffeeef32b9c950ywan      info.frame_height <= 0 ||
131233d2500723e5594f3e7c70896ffeeef32b9c950ywan      (info.frame_width % 2) != 0 ||
132233d2500723e5594f3e7c70896ffeeef32b9c950ywan      (info.frame_height % 2) != 0) {
133233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
134233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
135233d2500723e5594f3e7c70896ffeeef32b9c950ywan
136233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
137233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                             info.frame_height, 1)) {
138233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Failed to allocate image.");
139233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
140233d2500723e5594f3e7c70896ffeeef32b9c950ywan
141233d2500723e5594f3e7c70896ffeeef32b9c950ywan  printf("Using %s\n", vpx_codec_iface_name(encoder->interface()));
142233d2500723e5594f3e7c70896ffeeef32b9c950ywan
143233d2500723e5594f3e7c70896ffeeef32b9c950ywan  res = vpx_codec_enc_config_default(encoder->interface(), &cfg, 0);
144233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (res)
145233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(&codec, "Failed to get default codec config.");
146233d2500723e5594f3e7c70896ffeeef32b9c950ywan
147233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_w = info.frame_width;
148233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_h = info.frame_height;
149233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_timebase.num = info.time_base.numerator;
150233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_timebase.den = info.time_base.denominator;
151233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.rc_target_bitrate = bitrate;
152233d2500723e5594f3e7c70896ffeeef32b9c950ywan
153233d2500723e5594f3e7c70896ffeeef32b9c950ywan  writer = vpx_video_writer_open(argv[4], kContainerIVF, &info);
154233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!writer)
155233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Failed to open %s for writing.", argv[4]);
156233d2500723e5594f3e7c70896ffeeef32b9c950ywan
157233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!(infile = fopen(argv[3], "rb")))
158233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Failed to open %s for reading.", argv[3]);
159233d2500723e5594f3e7c70896ffeeef32b9c950ywan
160233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
161233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(&codec, "Failed to initialize encoder");
162233d2500723e5594f3e7c70896ffeeef32b9c950ywan
163233d2500723e5594f3e7c70896ffeeef32b9c950ywan  while (vpx_img_read(&raw, infile)) {
164233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (frame_count + 1 == update_frame_num) {
165233d2500723e5594f3e7c70896ffeeef32b9c950ywan      vpx_ref_frame_t ref;
166233d2500723e5594f3e7c70896ffeeef32b9c950ywan      ref.frame_type = VP8_LAST_FRAME;
167233d2500723e5594f3e7c70896ffeeef32b9c950ywan      ref.img = raw;
168233d2500723e5594f3e7c70896ffeeef32b9c950ywan      if (vpx_codec_control(&codec, VP8_SET_REFERENCE, &ref))
169233d2500723e5594f3e7c70896ffeeef32b9c950ywan        die_codec(&codec, "Failed to set reference frame");
170233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
171233d2500723e5594f3e7c70896ffeeef32b9c950ywan
172233d2500723e5594f3e7c70896ffeeef32b9c950ywan    encode_frame(&codec, &raw, frame_count++, writer);
173233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
174233d2500723e5594f3e7c70896ffeeef32b9c950ywan  encode_frame(&codec, NULL, -1, writer);
175233d2500723e5594f3e7c70896ffeeef32b9c950ywan
176233d2500723e5594f3e7c70896ffeeef32b9c950ywan  printf("\n");
177233d2500723e5594f3e7c70896ffeeef32b9c950ywan  fclose(infile);
178233d2500723e5594f3e7c70896ffeeef32b9c950ywan  printf("Processed %d frames.\n", frame_count);
179233d2500723e5594f3e7c70896ffeeef32b9c950ywan
180233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_img_free(&raw);
181233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (vpx_codec_destroy(&codec))
182233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(&codec, "Failed to destroy codec.");
183233d2500723e5594f3e7c70896ffeeef32b9c950ywan
184233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_video_writer_close(writer);
185233d2500723e5594f3e7c70896ffeeef32b9c950ywan
186233d2500723e5594f3e7c70896ffeeef32b9c950ywan  return EXIT_SUCCESS;
187233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
188