h264_parser: eliminate H264SliceContext usage
[libav.git] / libavcodec / h264_sei.c
CommitLineData
9c095463
MN
1/*
2 * H.26L/H.264/AVC/JVT/14496-10/... sei decoding
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
2912e87a 5 * This file is part of Libav.
9c095463 6 *
2912e87a 7 * Libav is free software; you can redistribute it and/or
9c095463
MN
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
2912e87a 12 * Libav is distributed in the hope that it will be useful,
9c095463
MN
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
2912e87a 18 * License along with Libav; if not, write to the Free Software
9c095463
MN
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
ba87f080 23 * @file
9c095463
MN
24 * H.264 / AVC / MPEG4 part10 sei decoding.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
9c095463 28#include "avcodec.h"
9c095463 29#include "golomb.h"
066ad249
VG
30#include "h264.h"
31#include "internal.h"
9c095463 32
b1841217
VG
33static const uint8_t sei_num_clock_ts_table[9] = {
34 1, 1, 1, 2, 2, 3, 3, 2, 3
9c095463
MN
35};
36
b1841217
VG
37void ff_h264_reset_sei(H264Context *h)
38{
9c095463
MN
39 h->sei_recovery_frame_cnt = -1;
40 h->sei_dpb_output_delay = 0;
41 h->sei_cpb_removal_delay = -1;
42 h->sei_buffering_period_present = 0;
5b10ef72 43 h->sei_frame_packing_present = 0;
18e3d61e 44 h->sei_display_orientation_present = 0;
271ce76d 45 h->sei_reguserdata_afd_present = 0;
b0017579
KK
46
47 h->a53_caption_size = 0;
48 av_freep(&h->a53_caption);
9c095463
MN
49}
50
b1841217
VG
51static int decode_picture_timing(H264Context *h)
52{
53 if (h->sps.nal_hrd_parameters_present_flag ||
54 h->sps.vcl_hrd_parameters_present_flag) {
55 h->sei_cpb_removal_delay = get_bits(&h->gb,
56 h->sps.cpb_removal_delay_length);
57 h->sei_dpb_output_delay = get_bits(&h->gb,
58 h->sps.dpb_output_delay_length);
9c095463 59 }
b1841217 60 if (h->sps.pic_struct_present_flag) {
9c095463 61 unsigned int i, num_clock_ts;
b1841217 62
2c541554 63 h->sei_pic_struct = get_bits(&h->gb, 4);
9c095463
MN
64 h->sei_ct_type = 0;
65
66 if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
0d8b943d 67 return AVERROR_INVALIDDATA;
9c095463
MN
68
69 num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
70
b1841217
VG
71 for (i = 0; i < num_clock_ts; i++) {
72 if (get_bits(&h->gb, 1)) { /* clock_timestamp_flag */
9c095463 73 unsigned int full_timestamp_flag;
b1841217
VG
74
75 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
2c541554
AK
76 skip_bits(&h->gb, 1); /* nuit_field_based_flag */
77 skip_bits(&h->gb, 5); /* counting_type */
78 full_timestamp_flag = get_bits(&h->gb, 1);
79 skip_bits(&h->gb, 1); /* discontinuity_flag */
80 skip_bits(&h->gb, 1); /* cnt_dropped_flag */
81 skip_bits(&h->gb, 8); /* n_frames */
b1841217 82 if (full_timestamp_flag) {
2c541554
AK
83 skip_bits(&h->gb, 6); /* seconds_value 0..59 */
84 skip_bits(&h->gb, 6); /* minutes_value 0..59 */
85 skip_bits(&h->gb, 5); /* hours_value 0..23 */
b1841217
VG
86 } else {
87 if (get_bits(&h->gb, 1)) { /* seconds_flag */
2c541554 88 skip_bits(&h->gb, 6); /* seconds_value range 0..59 */
b1841217 89 if (get_bits(&h->gb, 1)) { /* minutes_flag */
2c541554 90 skip_bits(&h->gb, 6); /* minutes_value 0..59 */
b1841217 91 if (get_bits(&h->gb, 1)) /* hours_flag */
2c541554 92 skip_bits(&h->gb, 5); /* hours_value 0..23 */
9c095463
MN
93 }
94 }
95 }
b1841217
VG
96 if (h->sps.time_offset_length > 0)
97 skip_bits(&h->gb,
98 h->sps.time_offset_length); /* time_offset */
9c095463
MN
99 }
100 }
101
b1841217
VG
102 if (h->avctx->debug & FF_DEBUG_PICT_INFO)
103 av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
104 h->sei_ct_type, h->sei_pic_struct);
9c095463
MN
105 }
106 return 0;
107}
108
80ea6611
DH
109static int decode_registered_user_data_afd(H264Context *h, int size)
110{
111 int flag;
112
113 if (size-- < 1)
114 return AVERROR_INVALIDDATA;
115 skip_bits(&h->gb, 1); // 0
116 flag = get_bits(&h->gb, 1); // active_format_flag
117 skip_bits(&h->gb, 6); // reserved
118
119 if (flag) {
120 if (size-- < 1)
121 return AVERROR_INVALIDDATA;
122 skip_bits(&h->gb, 4); // reserved
123 h->active_format_description = get_bits(&h->gb, 4);
124 h->sei_reguserdata_afd_present = 1;
125 }
126
127 return 0;
128}
129
130static int decode_registered_user_data_closed_caption(H264Context *h, int size)
131{
132 int flag;
133 int user_data_type_code;
134 int cc_count;
135
136 if (size < 3)
137 return AVERROR(EINVAL);
138
139 user_data_type_code = get_bits(&h->gb, 8);
140 if (user_data_type_code == 0x3) {
141 skip_bits(&h->gb, 1); // reserved
142
143 flag = get_bits(&h->gb, 1); // process_cc_data_flag
144 if (flag) {
145 skip_bits(&h->gb, 1); // zero bit
146 cc_count = get_bits(&h->gb, 5);
147 skip_bits(&h->gb, 8); // reserved
148 size -= 2;
149
150 if (cc_count && size >= cc_count * 3) {
151 const uint64_t new_size = (h->a53_caption_size + cc_count
152 * UINT64_C(3));
153 int i, ret;
154
155 if (new_size > INT_MAX)
156 return AVERROR(EINVAL);
157
158 /* Allow merging of the cc data from two fields. */
159 ret = av_reallocp(&h->a53_caption, new_size);
160 if (ret < 0)
161 return ret;
162
163 for (i = 0; i < cc_count; i++) {
164 h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
165 h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
166 h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
167 }
168
169 skip_bits(&h->gb, 8); // marker_bits
170 }
171 }
f9ab4fe1
JH
172 } else {
173 int i;
61d8fa2a 174 avpriv_request_sample(h->avctx, "Subtitles with data type 0x%02x",
f9ab4fe1
JH
175 user_data_type_code);
176 for (i = 0; i < size - 1; i++)
177 skip_bits(&h->gb, 8);
80ea6611
DH
178 }
179
180 return 0;
181}
182
271ce76d
VG
183static int decode_registered_user_data(H264Context *h, int size)
184{
185 uint32_t country_code;
186 uint32_t user_identifier;
271ce76d
VG
187
188 if (size < 7)
189 return AVERROR_INVALIDDATA;
190 size -= 7;
191
192 country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
193 if (country_code == 0xFF) {
194 skip_bits(&h->gb, 8); // itu_t_t35_country_code_extension_byte
195 size--;
196 }
197
198 /* itu_t_t35_payload_byte follows */
199 skip_bits(&h->gb, 8); // terminal provider code
200 skip_bits(&h->gb, 8); // terminal provider oriented code
201 user_identifier = get_bits_long(&h->gb, 32);
202
203 switch (user_identifier) {
204 case MKBETAG('D', 'T', 'G', '1'): // afd_data
80ea6611 205 return decode_registered_user_data_afd(h, size);
b0017579 206 case MKBETAG('G', 'A', '9', '4'): // closed captions
80ea6611 207 return decode_registered_user_data_closed_caption(h, size);
271ce76d
VG
208 default:
209 skip_bits(&h->gb, size * 8);
210 break;
211 }
212
213 return 0;
214}
215
b1841217
VG
216static int decode_unregistered_user_data(H264Context *h, int size)
217{
b4a82e74 218 uint8_t *user_data;
9c095463
MN
219 int e, build, i;
220
b4a82e74 221 if (size < 16 || size >= INT_MAX - 16)
0d8b943d 222 return AVERROR_INVALIDDATA;
9c095463 223
b4a82e74
VG
224 user_data = av_malloc(16 + size + 1);
225 if (!user_data)
226 return AVERROR(ENOMEM);
227
228 for (i = 0; i < size + 16; i++)
b1841217 229 user_data[i] = get_bits(&h->gb, 8);
9c095463 230
b1841217 231 user_data[i] = 0;
4c7fd58f 232 e = sscanf(user_data + 16, "x264 - core %d", &build);
b1841217
VG
233 if (e == 1 && build > 0)
234 h->x264_build = build;
9c095463 235
8487987b 236 if (strlen(user_data + 16) > 0)
b1841217 237 av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
9c095463 238
b4a82e74 239 av_free(user_data);
9c095463
MN
240 return 0;
241}
242
b1841217
VG
243static int decode_recovery_point(H264Context *h)
244{
2c541554 245 h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
b1841217
VG
246
247 /* 1b exact_match_flag,
248 * 1b broken_link_flag,
249 * 2b changing_slice_group_idc */
250 skip_bits(&h->gb, 4);
9c095463
MN
251
252 return 0;
253}
254
b1841217
VG
255static int decode_buffering_period(H264Context *h)
256{
9c095463
MN
257 unsigned int sps_id;
258 int sched_sel_idx;
259 SPS *sps;
260
2c541554 261 sps_id = get_ue_golomb_31(&h->gb);
b1841217
VG
262 if (sps_id > 31 || !h->sps_buffers[sps_id]) {
263 av_log(h->avctx, AV_LOG_ERROR,
264 "non-existing SPS %d referenced in buffering period\n", sps_id);
0d8b943d 265 return AVERROR_INVALIDDATA;
9c095463
MN
266 }
267 sps = h->sps_buffers[sps_id];
268
269 // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
270 if (sps->nal_hrd_parameters_present_flag) {
271 for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
b1841217
VG
272 h->initial_cpb_removal_delay[sched_sel_idx] =
273 get_bits(&h->gb, sps->initial_cpb_removal_delay_length);
274 // initial_cpb_removal_delay_offset
275 skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
9c095463
MN
276 }
277 }
278 if (sps->vcl_hrd_parameters_present_flag) {
279 for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
b1841217
VG
280 h->initial_cpb_removal_delay[sched_sel_idx] =
281 get_bits(&h->gb, sps->initial_cpb_removal_delay_length);
282 // initial_cpb_removal_delay_offset
283 skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
9c095463
MN
284 }
285 }
286
287 h->sei_buffering_period_present = 1;
288 return 0;
289}
290
5b10ef72
VG
291static int decode_frame_packing_arrangement(H264Context *h)
292{
5b10ef72 293 get_ue_golomb(&h->gb); // frame_packing_arrangement_id
3d9fc21e
VG
294 h->sei_frame_packing_present = !get_bits1(&h->gb);
295
296 if (h->sei_frame_packing_present) {
297 h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
298 h->quincunx_subsampling = get_bits1(&h->gb);
299 h->content_interpretation_type = get_bits(&h->gb, 6);
5b10ef72
VG
300
301 // the following skips: spatial_flipping_flag, frame0_flipped_flag,
302 // field_views_flag, current_frame_is_frame0_flag,
303 // frame0_self_contained_flag, frame1_self_contained_flag
304 skip_bits(&h->gb, 6);
305
3d9fc21e 306 if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
5b10ef72
VG
307 skip_bits(&h->gb, 16); // frame[01]_grid_position_[xy]
308 skip_bits(&h->gb, 8); // frame_packing_arrangement_reserved_byte
309 get_ue_golomb(&h->gb); // frame_packing_arrangement_repetition_period
310 }
311 skip_bits1(&h->gb); // frame_packing_arrangement_extension_flag
312
5b10ef72
VG
313 return 0;
314}
315
18e3d61e
VG
316static int decode_display_orientation(H264Context *h)
317{
318 h->sei_display_orientation_present = !get_bits1(&h->gb);
319
320 if (h->sei_display_orientation_present) {
321 h->sei_hflip = get_bits1(&h->gb); // hor_flip
322 h->sei_vflip = get_bits1(&h->gb); // ver_flip
323
324 h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
325 get_ue_golomb(&h->gb); // display_orientation_repetition_period
326 skip_bits1(&h->gb); // display_orientation_extension_flag
327 }
328
329 return 0;
330}
331
b1841217
VG
332int ff_h264_decode_sei(H264Context *h)
333{
2c541554 334 while (get_bits_left(&h->gb) > 16) {
b1841217
VG
335 int size = 0;
336 int type = 0;
0d8b943d 337 int ret = 0;
7ab551f9 338 int last = 0;
9c095463 339
7ab551f9
VG
340 while (get_bits_left(&h->gb) >= 8 &&
341 (last = get_bits(&h->gb, 8)) == 255) {
342 type += 255;
343 }
344 type += last;
9c095463 345
7ab551f9
VG
346 last = 0;
347 while (get_bits_left(&h->gb) >= 8 &&
348 (last = get_bits(&h->gb, 8)) == 255) {
349 size += 255;
350 }
351 size += last;
9c095463 352
84df6d6a
MN
353 if (size > get_bits_left(&h->gb) / 8) {
354 av_log(h->avctx, AV_LOG_ERROR, "SEI type %d truncated at %d\n",
355 type, get_bits_left(&h->gb));
356 return AVERROR_INVALIDDATA;
357 }
358
b1841217 359 switch (type) {
9c095463 360 case SEI_TYPE_PIC_TIMING: // Picture timing SEI
0d8b943d 361 ret = decode_picture_timing(h);
9c095463 362 break;
271ce76d
VG
363 case SEI_TYPE_USER_DATA_REGISTERED:
364 ret = decode_registered_user_data(h, size);
365 break;
9c095463 366 case SEI_TYPE_USER_DATA_UNREGISTERED:
0d8b943d 367 ret = decode_unregistered_user_data(h, size);
9c095463
MN
368 break;
369 case SEI_TYPE_RECOVERY_POINT:
0d8b943d 370 ret = decode_recovery_point(h);
9c095463 371 break;
304e916a 372 case SEI_TYPE_BUFFERING_PERIOD:
0d8b943d 373 ret = decode_buffering_period(h);
9c095463 374 break;
5b10ef72
VG
375 case SEI_TYPE_FRAME_PACKING:
376 ret = decode_frame_packing_arrangement(h);
5b10ef72 377 break;
18e3d61e
VG
378 case SEI_TYPE_DISPLAY_ORIENTATION:
379 ret = decode_display_orientation(h);
18e3d61e 380 break;
9c095463 381 default:
26179964 382 av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
b1841217 383 skip_bits(&h->gb, 8 * size);
9c095463 384 }
0bfab80a
VG
385 if (ret < 0)
386 return ret;
9c095463 387
b1841217 388 // FIXME check bits here
2c541554 389 align_get_bits(&h->gb);
9c095463
MN
390 }
391
392 return 0;
393}