h264_sei: Add namespace prefix to all SEI values
[libav.git] / libavcodec / h264_parser.c
1 /*
2 * H.26L/H.264/AVC/JVT/14496-10/... parser
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
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 *
12 * Libav is distributed in the hope that it will be useful,
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
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * H.264 / AVC / MPEG-4 part10 parser.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28 #include <assert.h>
29 #include <stdint.h>
30
31 #include "libavutil/avutil.h"
32 #include "libavutil/error.h"
33 #include "libavutil/log.h"
34 #include "libavutil/mem.h"
35 #include "libavutil/pixfmt.h"
36
37 #include "avcodec.h"
38 #include "get_bits.h"
39 #include "golomb_legacy.h"
40 #include "h264.h"
41 #include "h264_sei.h"
42 #include "h264_ps.h"
43 #include "h264data.h"
44 #include "internal.h"
45 #include "mpegutils.h"
46 #include "parser.h"
47
48 typedef struct H264ParseContext {
49 ParseContext pc;
50 H264ParamSets ps;
51 H264DSPContext h264dsp;
52 H264POCContext poc;
53 H264SEIContext sei;
54 int is_avc;
55 int nal_length_size;
56 int got_first;
57 int picture_structure;
58 } H264ParseContext;
59
60
61 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
62 int buf_size)
63 {
64 int i;
65 uint32_t state;
66 ParseContext *pc = &p->pc;
67 // mb_addr= pc->mb_addr - 1;
68 state = pc->state;
69 if (state > 13)
70 state = 7;
71
72 for (i = 0; i < buf_size; i++) {
73 if (state == 7) {
74 i += p->h264dsp.startcode_find_candidate(buf + i, buf_size - i);
75 if (i < buf_size)
76 state = 2;
77 } else if (state <= 2) {
78 if (buf[i] == 1)
79 state ^= 5; // 2->7, 1->4, 0->5
80 else if (buf[i])
81 state = 7;
82 else
83 state >>= 1; // 2->1, 1->0, 0->0
84 } else if (state <= 5) {
85 int nalu_type = buf[i] & 0x1F;
86 if (nalu_type == H264_NAL_SEI || nalu_type == H264_NAL_SPS ||
87 nalu_type == H264_NAL_PPS || nalu_type == H264_NAL_AUD) {
88 if (pc->frame_start_found) {
89 i++;
90 goto found;
91 }
92 } else if (nalu_type == H264_NAL_SLICE || nalu_type == H264_NAL_DPA ||
93 nalu_type == H264_NAL_IDR_SLICE) {
94 if (pc->frame_start_found) {
95 state += 8;
96 continue;
97 } else
98 pc->frame_start_found = 1;
99 }
100 state = 7;
101 } else {
102 // first_mb_in_slice is 0, probably the first nal of a new slice
103 if (buf[i] & 0x80)
104 goto found;
105 state = 7;
106 }
107 }
108 pc->state = state;
109 return END_NOT_FOUND;
110
111 found:
112 pc->state = 7;
113 pc->frame_start_found = 0;
114 return i - (state & 5);
115 }
116
117 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
118 AVCodecContext *avctx)
119 {
120 H264PredWeightTable pwt;
121 int slice_type_nos = s->pict_type & 3;
122 H264ParseContext *p = s->priv_data;
123 int list_count, ref_count[2];
124
125
126 if (p->ps.pps->redundant_pic_cnt_present)
127 get_ue_golomb(gb); // redundant_pic_count
128
129 if (slice_type_nos == AV_PICTURE_TYPE_B)
130 get_bits1(gb); // direct_spatial_mv_pred
131
132 if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
133 slice_type_nos, p->picture_structure) < 0)
134 return AVERROR_INVALIDDATA;
135
136 if (slice_type_nos != AV_PICTURE_TYPE_I) {
137 int list;
138 for (list = 0; list < list_count; list++) {
139 if (get_bits1(gb)) {
140 int index;
141 for (index = 0; ; index++) {
142 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
143
144 if (reordering_of_pic_nums_idc < 3)
145 get_ue_golomb(gb);
146 else if (reordering_of_pic_nums_idc > 3) {
147 av_log(avctx, AV_LOG_ERROR,
148 "illegal reordering_of_pic_nums_idc %d\n",
149 reordering_of_pic_nums_idc);
150 return AVERROR_INVALIDDATA;
151 } else
152 break;
153
154 if (index >= ref_count[list]) {
155 av_log(avctx, AV_LOG_ERROR,
156 "reference count %d overflow\n", index);
157 return AVERROR_INVALIDDATA;
158 }
159 }
160 }
161 }
162 }
163
164 if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
165 (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
166 ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
167 &pwt);
168
169 if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
170 int i;
171 for (i = 0; i < MAX_MMCO_COUNT; i++) {
172 MMCOOpcode opcode = get_ue_golomb_31(gb);
173 if (opcode > (unsigned) MMCO_LONG) {
174 av_log(avctx, AV_LOG_ERROR,
175 "illegal memory management control operation %d\n",
176 opcode);
177 return AVERROR_INVALIDDATA;
178 }
179 if (opcode == MMCO_END)
180 return 0;
181 else if (opcode == MMCO_RESET)
182 return 1;
183
184 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
185 get_ue_golomb(gb);
186 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
187 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
188 get_ue_golomb_31(gb);
189 }
190 }
191
192 return 0;
193 }
194
195 /**
196 * Parse NAL units of found picture and decode some basic information.
197 *
198 * @param s parser context.
199 * @param avctx codec context.
200 * @param buf buffer with field/frame data.
201 * @param buf_size size of the buffer.
202 */
203 static inline int parse_nal_units(AVCodecParserContext *s,
204 AVCodecContext *avctx,
205 const uint8_t *buf, int buf_size)
206 {
207 H264ParseContext *p = s->priv_data;
208 const uint8_t *buf_end = buf + buf_size;
209
210 H2645NAL nal = { NULL };
211
212 unsigned int pps_id;
213 unsigned int slice_type;
214 int state = -1, got_reset = 0;
215 int field_poc[2];
216 int ret;
217
218 /* set some sane default values */
219 s->pict_type = AV_PICTURE_TYPE_I;
220 s->key_frame = 0;
221 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
222
223 ff_h264_sei_uninit(&p->sei);
224
225 if (!buf_size)
226 return 0;
227
228 for (;;) {
229 const SPS *sps;
230 int src_length, consumed;
231 buf = avpriv_find_start_code(buf, buf_end, &state);
232 if (buf >= buf_end)
233 break;
234 --buf;
235 src_length = buf_end - buf;
236 switch (state & 0x1f) {
237 case H264_NAL_SLICE:
238 case H264_NAL_IDR_SLICE:
239 // Do not walk the whole buffer just to decode slice header
240 if ((state & 0x1f) == H264_NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
241 /* IDR or disposable slice
242 * No need to decode many bytes because MMCOs shall not be present. */
243 if (src_length > 60)
244 src_length = 60;
245 } else {
246 /* To decode up to MMCOs */
247 if (src_length > 1000)
248 src_length = 1000;
249 }
250 break;
251 }
252
253 consumed = ff_h2645_extract_rbsp(buf, src_length, &nal);
254 if (consumed < 0)
255 break;
256
257 ret = init_get_bits(&nal.gb, nal.data, nal.size * 8);
258 if (ret < 0)
259 goto fail;
260 get_bits1(&nal.gb);
261 nal.ref_idc = get_bits(&nal.gb, 2);
262 nal.type = get_bits(&nal.gb, 5);
263
264 switch (nal.type) {
265 case H264_NAL_SPS:
266 ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps);
267 break;
268 case H264_NAL_PPS:
269 ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
270 nal.size_bits);
271 break;
272 case H264_NAL_SEI:
273 ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
274 break;
275 case H264_NAL_IDR_SLICE:
276 s->key_frame = 1;
277
278 p->poc.prev_frame_num = 0;
279 p->poc.prev_frame_num_offset = 0;
280 p->poc.prev_poc_msb =
281 p->poc.prev_poc_lsb = 0;
282 /* fall through */
283 case H264_NAL_SLICE:
284 get_ue_golomb(&nal.gb); // skip first_mb_in_slice
285 slice_type = get_ue_golomb_31(&nal.gb);
286 s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
287 if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
288 /* key frame, since recovery_frame_cnt is set */
289 s->key_frame = 1;
290 }
291 pps_id = get_ue_golomb(&nal.gb);
292 if (pps_id >= MAX_PPS_COUNT) {
293 av_log(avctx, AV_LOG_ERROR,
294 "pps_id %u out of range\n", pps_id);
295 goto fail;
296 }
297 if (!p->ps.pps_list[pps_id]) {
298 av_log(avctx, AV_LOG_ERROR,
299 "non-existing PPS %u referenced\n", pps_id);
300 goto fail;
301 }
302 p->ps.pps = (const PPS*)p->ps.pps_list[pps_id]->data;
303 if (!p->ps.sps_list[p->ps.pps->sps_id]) {
304 av_log(avctx, AV_LOG_ERROR,
305 "non-existing SPS %u referenced\n", p->ps.pps->sps_id);
306 goto fail;
307 }
308 p->ps.sps = (SPS*)p->ps.sps_list[p->ps.pps->sps_id]->data;
309
310 sps = p->ps.sps;
311
312 p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
313
314 s->coded_width = 16 * sps->mb_width;
315 s->coded_height = 16 * sps->mb_height;
316 s->width = s->coded_width - (sps->crop_right + sps->crop_left);
317 s->height = s->coded_height - (sps->crop_top + sps->crop_bottom);
318 if (s->width <= 0 || s->height <= 0) {
319 s->width = s->coded_width;
320 s->height = s->coded_height;
321 }
322
323 switch (sps->bit_depth_luma) {
324 case 9:
325 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P9;
326 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
327 else s->format = AV_PIX_FMT_YUV420P9;
328 break;
329 case 10:
330 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P10;
331 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
332 else s->format = AV_PIX_FMT_YUV420P10;
333 break;
334 case 8:
335 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P;
336 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
337 else s->format = AV_PIX_FMT_YUV420P;
338 break;
339 default:
340 s->format = AV_PIX_FMT_NONE;
341 }
342
343 avctx->profile = ff_h264_get_profile(sps);
344 avctx->level = sps->level_idc;
345
346 if (sps->frame_mbs_only_flag) {
347 p->picture_structure = PICT_FRAME;
348 } else {
349 if (get_bits1(&nal.gb)) { // field_pic_flag
350 p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
351 } else {
352 p->picture_structure = PICT_FRAME;
353 }
354 }
355
356 if (nal.type == H264_NAL_IDR_SLICE)
357 get_ue_golomb(&nal.gb); /* idr_pic_id */
358 if (sps->poc_type == 0) {
359 p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
360
361 if (p->ps.pps->pic_order_present == 1 &&
362 p->picture_structure == PICT_FRAME)
363 p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
364 }
365
366 if (sps->poc_type == 1 &&
367 !sps->delta_pic_order_always_zero_flag) {
368 p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
369
370 if (p->ps.pps->pic_order_present == 1 &&
371 p->picture_structure == PICT_FRAME)
372 p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
373 }
374
375 /* Decode POC of this picture.
376 * The prev_ values needed for decoding POC of the next picture are not set here. */
377 field_poc[0] = field_poc[1] = INT_MAX;
378 ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
379 &p->poc, p->picture_structure, nal.ref_idc);
380
381 /* Continue parsing to check if MMCO_RESET is present.
382 * FIXME: MMCO_RESET could appear in non-first slice.
383 * Maybe, we should parse all undisposable non-IDR slice of this
384 * picture until encountering MMCO_RESET in a slice of it. */
385 if (nal.ref_idc && nal.type != H264_NAL_IDR_SLICE) {
386 got_reset = scan_mmco_reset(s, &nal.gb, avctx);
387 if (got_reset < 0)
388 goto fail;
389 }
390
391 /* Set up the prev_ values for decoding POC of the next picture. */
392 p->poc.prev_frame_num = got_reset ? 0 : p->poc.frame_num;
393 p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
394 if (nal.ref_idc != 0) {
395 if (!got_reset) {
396 p->poc.prev_poc_msb = p->poc.poc_msb;
397 p->poc.prev_poc_lsb = p->poc.poc_lsb;
398 } else {
399 p->poc.prev_poc_msb = 0;
400 p->poc.prev_poc_lsb =
401 p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
402 }
403 }
404
405 if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
406 switch (p->sei.picture_timing.pic_struct) {
407 case H264_SEI_PIC_STRUCT_TOP_FIELD:
408 case H264_SEI_PIC_STRUCT_BOTTOM_FIELD:
409 s->repeat_pict = 0;
410 break;
411 case H264_SEI_PIC_STRUCT_FRAME:
412 case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
413 case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
414 s->repeat_pict = 1;
415 break;
416 case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
417 case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
418 s->repeat_pict = 2;
419 break;
420 case H264_SEI_PIC_STRUCT_FRAME_DOUBLING:
421 s->repeat_pict = 3;
422 break;
423 case H264_SEI_PIC_STRUCT_FRAME_TRIPLING:
424 s->repeat_pict = 5;
425 break;
426 default:
427 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
428 break;
429 }
430 } else {
431 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
432 }
433
434 if (p->picture_structure == PICT_FRAME) {
435 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
436 if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
437 switch (p->sei.picture_timing.pic_struct) {
438 case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
439 case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
440 s->field_order = AV_FIELD_TT;
441 break;
442 case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
443 case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
444 s->field_order = AV_FIELD_BB;
445 break;
446 default:
447 s->field_order = AV_FIELD_PROGRESSIVE;
448 break;
449 }
450 } else {
451 if (field_poc[0] < field_poc[1])
452 s->field_order = AV_FIELD_TT;
453 else if (field_poc[0] > field_poc[1])
454 s->field_order = AV_FIELD_BB;
455 else
456 s->field_order = AV_FIELD_PROGRESSIVE;
457 }
458 } else {
459 if (p->picture_structure == PICT_TOP_FIELD)
460 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
461 else
462 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
463 s->field_order = AV_FIELD_UNKNOWN;
464 }
465
466 av_freep(&nal.rbsp_buffer);
467 return 0; /* no need to evaluate the rest */
468 }
469 buf += consumed;
470 }
471 /* didn't find a picture! */
472 av_log(avctx, AV_LOG_ERROR, "missing picture in access unit\n");
473 fail:
474 av_freep(&nal.rbsp_buffer);
475 return -1;
476 }
477
478 static int h264_parse(AVCodecParserContext *s,
479 AVCodecContext *avctx,
480 const uint8_t **poutbuf, int *poutbuf_size,
481 const uint8_t *buf, int buf_size)
482 {
483 H264ParseContext *p = s->priv_data;
484 ParseContext *pc = &p->pc;
485 int next;
486
487 if (!p->got_first) {
488 p->got_first = 1;
489 if (avctx->extradata_size) {
490 ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
491 &p->ps, &p->is_avc, &p->nal_length_size,
492 avctx->err_recognition, avctx);
493 }
494 }
495
496 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
497 next = buf_size;
498 } else {
499 next = h264_find_frame_end(p, buf, buf_size);
500
501 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
502 *poutbuf = NULL;
503 *poutbuf_size = 0;
504 return buf_size;
505 }
506
507 if (next < 0 && next != END_NOT_FOUND) {
508 assert(pc->last_index + next >= 0);
509 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next); // update state
510 }
511 }
512
513 parse_nal_units(s, avctx, buf, buf_size);
514
515 if (p->sei.picture_timing.cpb_removal_delay >= 0) {
516 s->dts_sync_point = p->sei.buffering_period.present;
517 s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
518 s->pts_dts_delta = p->sei.picture_timing.dpb_output_delay;
519 } else {
520 s->dts_sync_point = INT_MIN;
521 s->dts_ref_dts_delta = INT_MIN;
522 s->pts_dts_delta = INT_MIN;
523 }
524
525 if (s->flags & PARSER_FLAG_ONCE) {
526 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
527 }
528
529 *poutbuf = buf;
530 *poutbuf_size = buf_size;
531 return next;
532 }
533
534 static int h264_split(AVCodecContext *avctx,
535 const uint8_t *buf, int buf_size)
536 {
537 int i;
538 uint32_t state = -1;
539 int has_sps = 0;
540
541 for (i = 0; i <= buf_size; i++) {
542 if ((state & 0xFFFFFF1F) == 0x107)
543 has_sps = 1;
544 /* if((state&0xFFFFFF1F) == 0x101 ||
545 * (state&0xFFFFFF1F) == 0x102 ||
546 * (state&0xFFFFFF1F) == 0x105) {
547 * }
548 */
549 if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x106 &&
550 (state & 0xFFFFFF1F) != 0x107 && (state & 0xFFFFFF1F) != 0x108 &&
551 (state & 0xFFFFFF1F) != 0x109 && (state & 0xFFFFFF1F) != 0x10d &&
552 (state & 0xFFFFFF1F) != 0x10f) {
553 if (has_sps) {
554 while (i > 4 && buf[i - 5] == 0)
555 i--;
556 return i - 4;
557 }
558 }
559 if (i < buf_size)
560 state = (state << 8) | buf[i];
561 }
562 return 0;
563 }
564
565 static void h264_close(AVCodecParserContext *s)
566 {
567 H264ParseContext *p = s->priv_data;
568 ParseContext *pc = &p->pc;
569 int i;
570
571 av_free(pc->buffer);
572
573 ff_h264_sei_uninit(&p->sei);
574
575 for (i = 0; i < FF_ARRAY_ELEMS(p->ps.sps_list); i++)
576 av_buffer_unref(&p->ps.sps_list[i]);
577
578 for (i = 0; i < FF_ARRAY_ELEMS(p->ps.pps_list); i++)
579 av_buffer_unref(&p->ps.pps_list[i]);
580 }
581
582 static av_cold int init(AVCodecParserContext *s)
583 {
584 H264ParseContext *p = s->priv_data;
585
586 ff_h264dsp_init(&p->h264dsp, 8, 1);
587 return 0;
588 }
589
590 AVCodecParser ff_h264_parser = {
591 .codec_ids = { AV_CODEC_ID_H264 },
592 .priv_data_size = sizeof(H264ParseContext),
593 .parser_init = init,
594 .parser_parse = h264_parse,
595 .parser_close = h264_close,
596 .split = h264_split,
597 };