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 // VP8 Set Active and ROI Maps
12 // ===========================
13 //
14 // This is an example demonstrating how to control the VP8 encoder's
15 // ROI and Active maps.
16 //
17 // ROI (Reigon of Interest) maps are a way for the application to assign
18 // each macroblock in the image to a region, and then set quantizer and
19 // filtering parameters on that image.
20 //
21 // Active maps are a way for the application to specify on a
22 // macroblock-by-macroblock basis whether there is any activity in that
23 // macroblock.
24 //
25 //
26 // Configuration
27 // -------------
28 // An ROI map is set on frame 22. If the width of the image in macroblocks
29 // is evenly divisble by 4, then the output will appear to have distinct
30 // columns, where the quantizer, loopfilter, and static threshold differ
31 // from column to column.
32 //
33 // An active map is set on frame 33. If the width of the image in macroblocks
34 // is evenly divisble by 4, then the output will appear to have distinct
35 // columns, where one column will have motion and the next will not.
36 //
37 // The active map is cleared on frame 44.
38 //
39 // Observing The Effects
40 // ---------------------
41 // Use the `simple_decoder` example to decode this sample, and observe
42 // the change in the image at frames 22, 33, and 44.
43
44 #include <assert.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48
49 #include "vpx/vp8cx.h"
50 #include "vpx/vpx_encoder.h"
51
52 #include "../tools_common.h"
53 #include "../video_writer.h"
54
55 static const char *exec_name;
56
usage_exit(void)57 void usage_exit(void) {
58 fprintf(stderr, "Usage: %s <codec> <width> <height> <infile> <outfile>\n",
59 exec_name);
60 exit(EXIT_FAILURE);
61 }
62
set_roi_map(const vpx_codec_enc_cfg_t * cfg,vpx_codec_ctx_t * codec)63 static void set_roi_map(const vpx_codec_enc_cfg_t *cfg,
64 vpx_codec_ctx_t *codec) {
65 unsigned int i;
66 vpx_roi_map_t roi;
67 memset(&roi, 0, sizeof(roi));
68
69 roi.rows = (cfg->g_h + 15) / 16;
70 roi.cols = (cfg->g_w + 15) / 16;
71
72 roi.delta_q[0] = 0;
73 roi.delta_q[1] = -2;
74 roi.delta_q[2] = -4;
75 roi.delta_q[3] = -6;
76
77 roi.delta_lf[0] = 0;
78 roi.delta_lf[1] = 1;
79 roi.delta_lf[2] = 2;
80 roi.delta_lf[3] = 3;
81
82 roi.static_threshold[0] = 1500;
83 roi.static_threshold[1] = 1000;
84 roi.static_threshold[2] = 500;
85 roi.static_threshold[3] = 0;
86
87 roi.roi_map = (uint8_t *)malloc(roi.rows * roi.cols);
88 for (i = 0; i < roi.rows * roi.cols; ++i) roi.roi_map[i] = i % 4;
89
90 if (vpx_codec_control(codec, VP8E_SET_ROI_MAP, &roi))
91 die_codec(codec, "Failed to set ROI map");
92
93 free(roi.roi_map);
94 }
95
set_active_map(const vpx_codec_enc_cfg_t * cfg,vpx_codec_ctx_t * codec)96 static void set_active_map(const vpx_codec_enc_cfg_t *cfg,
97 vpx_codec_ctx_t *codec) {
98 unsigned int i;
99 vpx_active_map_t map = { 0, 0, 0 };
100
101 map.rows = (cfg->g_h + 15) / 16;
102 map.cols = (cfg->g_w + 15) / 16;
103
104 map.active_map = (uint8_t *)malloc(map.rows * map.cols);
105 for (i = 0; i < map.rows * map.cols; ++i) map.active_map[i] = i % 2;
106
107 if (vpx_codec_control(codec, VP8E_SET_ACTIVEMAP, &map))
108 die_codec(codec, "Failed to set active map");
109
110 free(map.active_map);
111 }
112
unset_active_map(const vpx_codec_enc_cfg_t * cfg,vpx_codec_ctx_t * codec)113 static void unset_active_map(const vpx_codec_enc_cfg_t *cfg,
114 vpx_codec_ctx_t *codec) {
115 vpx_active_map_t map = { 0, 0, 0 };
116
117 map.rows = (cfg->g_h + 15) / 16;
118 map.cols = (cfg->g_w + 15) / 16;
119 map.active_map = NULL;
120
121 if (vpx_codec_control(codec, VP8E_SET_ACTIVEMAP, &map))
122 die_codec(codec, "Failed to set active map");
123 }
124
encode_frame(vpx_codec_ctx_t * codec,vpx_image_t * img,int frame_index,VpxVideoWriter * writer)125 static int encode_frame(vpx_codec_ctx_t *codec, vpx_image_t *img,
126 int frame_index, VpxVideoWriter *writer) {
127 int got_pkts = 0;
128 vpx_codec_iter_t iter = NULL;
129 const vpx_codec_cx_pkt_t *pkt = NULL;
130 const vpx_codec_err_t res =
131 vpx_codec_encode(codec, img, frame_index, 1, 0, VPX_DL_GOOD_QUALITY);
132 if (res != VPX_CODEC_OK) die_codec(codec, "Failed to encode frame");
133
134 while ((pkt = vpx_codec_get_cx_data(codec, &iter)) != NULL) {
135 got_pkts = 1;
136
137 if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) {
138 const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0;
139 if (!vpx_video_writer_write_frame(writer, pkt->data.frame.buf,
140 pkt->data.frame.sz,
141 pkt->data.frame.pts)) {
142 die_codec(codec, "Failed to write compressed frame");
143 }
144
145 printf(keyframe ? "K" : ".");
146 fflush(stdout);
147 }
148 }
149
150 return got_pkts;
151 }
152
main(int argc,char ** argv)153 int main(int argc, char **argv) {
154 FILE *infile = NULL;
155 vpx_codec_ctx_t codec;
156 vpx_codec_enc_cfg_t cfg;
157 int frame_count = 0;
158 vpx_image_t raw;
159 vpx_codec_err_t res;
160 VpxVideoInfo info;
161 VpxVideoWriter *writer = NULL;
162 const VpxInterface *encoder = NULL;
163 const int fps = 2; // TODO(dkovalev) add command line argument
164 const double bits_per_pixel_per_frame = 0.067;
165
166 exec_name = argv[0];
167 if (argc != 6) die("Invalid number of arguments");
168
169 memset(&info, 0, sizeof(info));
170
171 encoder = get_vpx_encoder_by_name(argv[1]);
172 if (encoder == NULL) {
173 die("Unsupported codec.");
174 }
175 assert(encoder != NULL);
176 info.codec_fourcc = encoder->fourcc;
177 info.frame_width = (int)strtol(argv[2], NULL, 0);
178 info.frame_height = (int)strtol(argv[3], NULL, 0);
179 info.time_base.numerator = 1;
180 info.time_base.denominator = fps;
181
182 if (info.frame_width <= 0 || info.frame_height <= 0 ||
183 (info.frame_width % 2) != 0 || (info.frame_height % 2) != 0) {
184 die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
185 }
186
187 if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
188 info.frame_height, 1)) {
189 die("Failed to allocate image.");
190 }
191
192 printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
193
194 res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
195 if (res) die_codec(&codec, "Failed to get default codec config.");
196
197 cfg.g_w = info.frame_width;
198 cfg.g_h = info.frame_height;
199 cfg.g_timebase.num = info.time_base.numerator;
200 cfg.g_timebase.den = info.time_base.denominator;
201 cfg.rc_target_bitrate =
202 (unsigned int)(bits_per_pixel_per_frame * cfg.g_w * cfg.g_h * fps / 1000);
203 cfg.g_lag_in_frames = 0;
204
205 writer = vpx_video_writer_open(argv[5], kContainerIVF, &info);
206 if (!writer) die("Failed to open %s for writing.", argv[5]);
207
208 if (!(infile = fopen(argv[4], "rb")))
209 die("Failed to open %s for reading.", argv[4]);
210
211 if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
212 die_codec(&codec, "Failed to initialize encoder");
213
214 // Encode frames.
215 while (vpx_img_read(&raw, infile)) {
216 ++frame_count;
217
218 if (frame_count == 22 && encoder->fourcc == VP8_FOURCC) {
219 set_roi_map(&cfg, &codec);
220 } else if (frame_count == 33) {
221 set_active_map(&cfg, &codec);
222 } else if (frame_count == 44) {
223 unset_active_map(&cfg, &codec);
224 }
225
226 encode_frame(&codec, &raw, frame_count, writer);
227 }
228
229 // Flush encoder.
230 while (encode_frame(&codec, NULL, -1, writer)) {
231 }
232
233 printf("\n");
234 fclose(infile);
235 printf("Processed %d frames.\n", frame_count);
236
237 vpx_img_free(&raw);
238 if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec.");
239
240 vpx_video_writer_close(writer);
241
242 return EXIT_SUCCESS;
243 }
244