2 * H.26L/H.264/AVC/JVT/14496-10/... parser
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of FFmpeg.
7 * FFmpeg 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.
12 * FFmpeg 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * H.264 / AVC / MPEG-4 part10 parser.
25 * @author Michael Niedermayer <michaelni@gmx.at>
28 #define UNCHECKED_BITSTREAM_READER 1
33 #include "libavutil/avutil.h"
34 #include "libavutil/error.h"
35 #include "libavutil/log.h"
36 #include "libavutil/mem.h"
37 #include "libavutil/pixfmt.h"
45 #include "mpegutils.h"
48 typedef struct H264ParseContext {
51 H264DSPContext h264dsp;
57 int picture_structure;
58 uint8_t parse_history[6];
59 int parse_history_count;
64 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
65 int buf_size, void *logctx)
69 ParseContext *pc = &p->pc;
71 int next_avc = p->is_avc ? 0 : buf_size;
72 // mb_addr= pc->mb_addr - 1;
77 if (p->is_avc && !p->nal_length_size)
78 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
80 for (i = 0; i < buf_size; i++) {
84 for (j = 0; j < p->nal_length_size; j++)
85 nalsize = (nalsize << 8) | buf[i++];
86 if (nalsize <= 0 || nalsize > buf_size - i) {
87 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
90 next_avc = i + nalsize;
95 i += p->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
98 } else if (state <= 2) {
100 state ^= 5; // 2->7, 1->4, 0->5
104 state >>= 1; // 2->1, 1->0, 0->0
105 } else if (state <= 5) {
106 int nalu_type = buf[i] & 0x1F;
107 if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
108 nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
109 if (pc->frame_start_found) {
113 } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
114 nalu_type == NAL_IDR_SLICE) {
120 p->parse_history[p->parse_history_count++] = buf[i];
121 if (p->parse_history_count > 5) {
122 unsigned int mb, last_mb = p->parse_last_mb;
125 init_get_bits(&gb, p->parse_history, 8*p->parse_history_count);
126 p->parse_history_count = 0;
127 mb= get_ue_golomb_long(&gb);
128 p->parse_last_mb = mb;
129 if (pc->frame_start_found) {
133 pc->frame_start_found = 1;
141 return END_NOT_FOUND;
145 pc->frame_start_found = 0;
148 return i - (state & 5) - 5 * (state > 7);
151 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
154 H264PredWeightTable pwt;
155 int slice_type_nos = s->pict_type & 3;
156 H264ParseContext *p = s->priv_data;
157 int list_count, ref_count[2];
160 if (p->ps.pps->redundant_pic_cnt_present)
161 get_ue_golomb(gb); // redundant_pic_count
163 if (slice_type_nos == AV_PICTURE_TYPE_B)
164 get_bits1(gb); // direct_spatial_mv_pred
166 if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
167 slice_type_nos, p->picture_structure, logctx) < 0)
168 return AVERROR_INVALIDDATA;
170 if (slice_type_nos != AV_PICTURE_TYPE_I) {
172 for (list = 0; list < list_count; list++) {
175 for (index = 0; ; index++) {
176 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
178 if (reordering_of_pic_nums_idc < 3)
179 get_ue_golomb_long(gb);
180 else if (reordering_of_pic_nums_idc > 3) {
181 av_log(logctx, AV_LOG_ERROR,
182 "illegal reordering_of_pic_nums_idc %d\n",
183 reordering_of_pic_nums_idc);
184 return AVERROR_INVALIDDATA;
188 if (index >= ref_count[list]) {
189 av_log(logctx, AV_LOG_ERROR,
190 "reference count %d overflow\n", index);
191 return AVERROR_INVALIDDATA;
198 if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
199 (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
200 ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
203 if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
205 for (i = 0; i < MAX_MMCO_COUNT; i++) {
206 MMCOOpcode opcode = get_ue_golomb_31(gb);
207 if (opcode > (unsigned) MMCO_LONG) {
208 av_log(logctx, AV_LOG_ERROR,
209 "illegal memory management control operation %d\n",
211 return AVERROR_INVALIDDATA;
213 if (opcode == MMCO_END)
215 else if (opcode == MMCO_RESET)
218 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
219 get_ue_golomb_long(gb); // difference_of_pic_nums_minus1
220 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
221 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
222 get_ue_golomb_31(gb);
229 static inline int get_avc_nalsize(H264ParseContext *p, const uint8_t *buf,
230 int buf_size, int *buf_index, void *logctx)
234 if (*buf_index >= buf_size - p->nal_length_size) {
235 // the end of the buffer is reached, refill it
236 return AVERROR(EAGAIN);
239 for (i = 0; i < p->nal_length_size; i++)
240 nalsize = ((unsigned)nalsize << 8) | buf[(*buf_index)++];
241 if (nalsize <= 0 || nalsize > buf_size - *buf_index) {
242 av_log(logctx, AV_LOG_ERROR,
243 "AVC: nal size %d\n", nalsize);
244 return AVERROR_INVALIDDATA;
250 * Parse NAL units of found picture and decode some basic information.
252 * @param s parser context.
253 * @param avctx codec context.
254 * @param buf buffer with field/frame data.
255 * @param buf_size size of the buffer.
257 static inline int parse_nal_units(AVCodecParserContext *s,
258 AVCodecContext *avctx,
259 const uint8_t * const buf, int buf_size)
261 H264ParseContext *p = s->priv_data;
262 H2645NAL nal = { NULL };
263 int buf_index, next_avc;
265 unsigned int slice_type;
266 int state = -1, got_reset = 0;
267 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
271 /* set some sane default values */
272 s->pict_type = AV_PICTURE_TYPE_I;
274 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
276 ff_h264_sei_uninit(&p->sei);
277 p->sei.frame_packing.frame_packing_arrangement_cancel_flag = -1;
283 next_avc = p->is_avc ? 0 : buf_size;
286 int src_length, consumed, nalsize = 0;
288 if (buf_index >= next_avc) {
289 nalsize = get_avc_nalsize(p, buf, buf_size, &buf_index, avctx);
292 next_avc = buf_index + nalsize;
294 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
295 if (buf_index >= buf_size)
297 if (buf_index >= next_avc)
300 src_length = next_avc - buf_index;
302 state = buf[buf_index];
303 switch (state & 0x1f) {
306 // Do not walk the whole buffer just to decode slice header
307 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
308 /* IDR or disposable slice
309 * No need to decode many bytes because MMCOs shall not be present. */
313 /* To decode up to MMCOs */
314 if (src_length > 1000)
319 consumed = ff_h2645_extract_rbsp(buf + buf_index, src_length, &nal);
323 buf_index += consumed;
325 ret = init_get_bits8(&nal.gb, nal.data, nal.size);
329 nal.ref_idc = get_bits(&nal.gb, 2);
330 nal.type = get_bits(&nal.gb, 5);
334 ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps, 0);
337 ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
341 ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
346 p->poc.prev_frame_num = 0;
347 p->poc.prev_frame_num_offset = 0;
348 p->poc.prev_poc_msb =
349 p->poc.prev_poc_lsb = 0;
352 get_ue_golomb_long(&nal.gb); // skip first_mb_in_slice
353 slice_type = get_ue_golomb_31(&nal.gb);
354 s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
355 if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
356 /* key frame, since recovery_frame_cnt is set */
359 pps_id = get_ue_golomb(&nal.gb);
360 if (pps_id >= MAX_PPS_COUNT) {
361 av_log(avctx, AV_LOG_ERROR,
362 "pps_id %u out of range\n", pps_id);
365 if (!p->ps.pps_list[pps_id]) {
366 av_log(avctx, AV_LOG_ERROR,
367 "non-existing PPS %u referenced\n", pps_id);
370 p->ps.pps = (const PPS*)p->ps.pps_list[pps_id]->data;
371 if (!p->ps.sps_list[p->ps.pps->sps_id]) {
372 av_log(avctx, AV_LOG_ERROR,
373 "non-existing SPS %u referenced\n", p->ps.pps->sps_id);
376 p->ps.sps = (SPS*)p->ps.sps_list[p->ps.pps->sps_id]->data;
380 // heuristic to detect non marked keyframes
381 if (p->ps.sps->ref_frame_count <= 1 && p->ps.pps->ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
384 p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
386 s->coded_width = 16 * sps->mb_width;
387 s->coded_height = 16 * sps->mb_height;
388 s->width = s->coded_width - (sps->crop_right + sps->crop_left);
389 s->height = s->coded_height - (sps->crop_top + sps->crop_bottom);
390 if (s->width <= 0 || s->height <= 0) {
391 s->width = s->coded_width;
392 s->height = s->coded_height;
395 switch (sps->bit_depth_luma) {
397 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P9;
398 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
399 else s->format = AV_PIX_FMT_YUV420P9;
402 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P10;
403 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
404 else s->format = AV_PIX_FMT_YUV420P10;
407 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P;
408 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
409 else s->format = AV_PIX_FMT_YUV420P;
412 s->format = AV_PIX_FMT_NONE;
415 avctx->profile = ff_h264_get_profile(sps);
416 avctx->level = sps->level_idc;
418 if (sps->frame_mbs_only_flag) {
419 p->picture_structure = PICT_FRAME;
421 if (get_bits1(&nal.gb)) { // field_pic_flag
422 p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
424 p->picture_structure = PICT_FRAME;
428 if (nal.type == NAL_IDR_SLICE)
429 get_ue_golomb_long(&nal.gb); /* idr_pic_id */
430 if (sps->poc_type == 0) {
431 p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
433 if (p->ps.pps->pic_order_present == 1 &&
434 p->picture_structure == PICT_FRAME)
435 p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
438 if (sps->poc_type == 1 &&
439 !sps->delta_pic_order_always_zero_flag) {
440 p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
442 if (p->ps.pps->pic_order_present == 1 &&
443 p->picture_structure == PICT_FRAME)
444 p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
447 /* Decode POC of this picture.
448 * The prev_ values needed for decoding POC of the next picture are not set here. */
449 field_poc[0] = field_poc[1] = INT_MAX;
450 ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
451 &p->poc, p->picture_structure, nal.ref_idc);
453 /* Continue parsing to check if MMCO_RESET is present.
454 * FIXME: MMCO_RESET could appear in non-first slice.
455 * Maybe, we should parse all undisposable non-IDR slice of this
456 * picture until encountering MMCO_RESET in a slice of it. */
457 if (nal.ref_idc && nal.type != NAL_IDR_SLICE) {
458 got_reset = scan_mmco_reset(s, &nal.gb, avctx);
463 /* Set up the prev_ values for decoding POC of the next picture. */
464 p->poc.prev_frame_num = got_reset ? 0 : p->poc.frame_num;
465 p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
466 if (nal.ref_idc != 0) {
468 p->poc.prev_poc_msb = p->poc.poc_msb;
469 p->poc.prev_poc_lsb = p->poc.poc_lsb;
471 p->poc.prev_poc_msb = 0;
472 p->poc.prev_poc_lsb =
473 p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
477 if (sps->pic_struct_present_flag) {
478 switch (p->sei.picture_timing.pic_struct) {
479 case SEI_PIC_STRUCT_TOP_FIELD:
480 case SEI_PIC_STRUCT_BOTTOM_FIELD:
483 case SEI_PIC_STRUCT_FRAME:
484 case SEI_PIC_STRUCT_TOP_BOTTOM:
485 case SEI_PIC_STRUCT_BOTTOM_TOP:
488 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
489 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
492 case SEI_PIC_STRUCT_FRAME_DOUBLING:
495 case SEI_PIC_STRUCT_FRAME_TRIPLING:
499 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
503 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
506 if (p->picture_structure == PICT_FRAME) {
507 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
508 if (sps->pic_struct_present_flag) {
509 switch (p->sei.picture_timing.pic_struct) {
510 case SEI_PIC_STRUCT_TOP_BOTTOM:
511 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
512 s->field_order = AV_FIELD_TT;
514 case SEI_PIC_STRUCT_BOTTOM_TOP:
515 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
516 s->field_order = AV_FIELD_BB;
519 s->field_order = AV_FIELD_PROGRESSIVE;
523 if (field_poc[0] < field_poc[1])
524 s->field_order = AV_FIELD_TT;
525 else if (field_poc[0] > field_poc[1])
526 s->field_order = AV_FIELD_BB;
528 s->field_order = AV_FIELD_PROGRESSIVE;
531 if (p->picture_structure == PICT_TOP_FIELD)
532 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
534 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
535 s->field_order = AV_FIELD_UNKNOWN;
538 av_freep(&nal.rbsp_buffer);
539 return 0; /* no need to evaluate the rest */
543 av_freep(&nal.rbsp_buffer);
546 /* didn't find a picture! */
547 av_log(avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
549 av_freep(&nal.rbsp_buffer);
553 static int h264_parse(AVCodecParserContext *s,
554 AVCodecContext *avctx,
555 const uint8_t **poutbuf, int *poutbuf_size,
556 const uint8_t *buf, int buf_size)
558 H264ParseContext *p = s->priv_data;
559 ParseContext *pc = &p->pc;
564 if (avctx->extradata_size) {
565 ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
566 &p->ps, &p->is_avc, &p->nal_length_size,
567 avctx->err_recognition, avctx);
571 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
574 next = h264_find_frame_end(p, buf, buf_size, avctx);
576 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
582 if (next < 0 && next != END_NOT_FOUND) {
583 av_assert1(pc->last_index + next >= 0);
584 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next, avctx); // update state
588 parse_nal_units(s, avctx, buf, buf_size);
590 if (avctx->framerate.num)
591 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
592 if (p->sei.picture_timing.cpb_removal_delay >= 0) {
593 s->dts_sync_point = p->sei.buffering_period.present;
594 s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
595 s->pts_dts_delta = p->sei.picture_timing.dpb_output_delay;
597 s->dts_sync_point = INT_MIN;
598 s->dts_ref_dts_delta = INT_MIN;
599 s->pts_dts_delta = INT_MIN;
602 if (s->flags & PARSER_FLAG_ONCE) {
603 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
607 *poutbuf_size = buf_size;
611 static int h264_split(AVCodecContext *avctx,
612 const uint8_t *buf, int buf_size)
617 const uint8_t *ptr = buf, *end = buf + buf_size;
621 ptr = avpriv_find_start_code(ptr, end, &state);
622 if ((state & 0xFFFFFF00) != 0x100)
624 nalu_type = state & 0x1F;
625 if (nalu_type == NAL_SPS) {
627 } else if (nalu_type == NAL_PPS)
629 /* else if (nalu_type == 0x01 ||
630 * nalu_type == 0x02 ||
631 * nalu_type == 0x05) {
634 else if ((nalu_type != NAL_SEI || has_pps) &&
635 nalu_type != NAL_AUD && nalu_type != NAL_SPS_EXT &&
638 while (ptr - 4 > buf && ptr[-5] == 0)
640 return ptr - 4 - buf;
648 static void h264_close(AVCodecParserContext *s)
650 H264ParseContext *p = s->priv_data;
651 ParseContext *pc = &p->pc;
653 av_freep(&pc->buffer);
655 ff_h264_sei_uninit(&p->sei);
656 ff_h264_ps_uninit(&p->ps);
659 static av_cold int init(AVCodecParserContext *s)
661 H264ParseContext *p = s->priv_data;
663 ff_h264dsp_init(&p->h264dsp, 8, 1);
667 AVCodecParser ff_h264_parser = {
668 .codec_ids = { AV_CODEC_ID_H264 },
669 .priv_data_size = sizeof(H264ParseContext),
671 .parser_parse = h264_parse,
672 .parser_close = h264_close,