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"
47 #include "mpegutils.h"
50 typedef struct H264ParseContext {
53 H264DSPContext h264dsp;
59 int picture_structure;
60 uint8_t parse_history[6];
61 int parse_history_count;
63 int64_t reference_dts;
67 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
68 int buf_size, void *logctx)
72 ParseContext *pc = &p->pc;
74 int next_avc = p->is_avc ? 0 : buf_size;
75 // mb_addr= pc->mb_addr - 1;
80 if (p->is_avc && !p->nal_length_size)
81 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
83 for (i = 0; i < buf_size; i++) {
87 for (j = 0; j < p->nal_length_size; j++)
88 nalsize = (nalsize << 8) | buf[i++];
89 if (nalsize <= 0 || nalsize > buf_size - i) {
90 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
93 next_avc = i + nalsize;
98 i += p->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
101 } else if (state <= 2) {
103 state ^= 5; // 2->7, 1->4, 0->5
107 state >>= 1; // 2->1, 1->0, 0->0
108 } else if (state <= 5) {
109 int nalu_type = buf[i] & 0x1F;
110 if (nalu_type == H264_NAL_SEI || nalu_type == H264_NAL_SPS ||
111 nalu_type == H264_NAL_PPS || nalu_type == H264_NAL_AUD) {
112 if (pc->frame_start_found) {
116 } else if (nalu_type == H264_NAL_SLICE || nalu_type == H264_NAL_DPA ||
117 nalu_type == H264_NAL_IDR_SLICE) {
123 p->parse_history[p->parse_history_count++] = buf[i];
124 if (p->parse_history_count > 5) {
125 unsigned int mb, last_mb = p->parse_last_mb;
128 init_get_bits(&gb, p->parse_history, 8*p->parse_history_count);
129 p->parse_history_count = 0;
130 mb= get_ue_golomb_long(&gb);
131 p->parse_last_mb = mb;
132 if (pc->frame_start_found) {
136 pc->frame_start_found = 1;
144 return END_NOT_FOUND;
148 pc->frame_start_found = 0;
151 return i - (state & 5) - 5 * (state > 7);
154 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
157 H264PredWeightTable pwt;
158 int slice_type_nos = s->pict_type & 3;
159 H264ParseContext *p = s->priv_data;
160 int list_count, ref_count[2];
163 if (p->ps.pps->redundant_pic_cnt_present)
164 get_ue_golomb(gb); // redundant_pic_count
166 if (slice_type_nos == AV_PICTURE_TYPE_B)
167 get_bits1(gb); // direct_spatial_mv_pred
169 if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
170 slice_type_nos, p->picture_structure, logctx) < 0)
171 return AVERROR_INVALIDDATA;
173 if (slice_type_nos != AV_PICTURE_TYPE_I) {
175 for (list = 0; list < list_count; list++) {
178 for (index = 0; ; index++) {
179 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
181 if (reordering_of_pic_nums_idc < 3)
182 get_ue_golomb_long(gb);
183 else if (reordering_of_pic_nums_idc > 3) {
184 av_log(logctx, AV_LOG_ERROR,
185 "illegal reordering_of_pic_nums_idc %d\n",
186 reordering_of_pic_nums_idc);
187 return AVERROR_INVALIDDATA;
191 if (index >= ref_count[list]) {
192 av_log(logctx, AV_LOG_ERROR,
193 "reference count %d overflow\n", index);
194 return AVERROR_INVALIDDATA;
201 if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
202 (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
203 ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
206 if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
208 for (i = 0; i < MAX_MMCO_COUNT; i++) {
209 MMCOOpcode opcode = get_ue_golomb_31(gb);
210 if (opcode > (unsigned) MMCO_LONG) {
211 av_log(logctx, AV_LOG_ERROR,
212 "illegal memory management control operation %d\n",
214 return AVERROR_INVALIDDATA;
216 if (opcode == MMCO_END)
218 else if (opcode == MMCO_RESET)
221 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
222 get_ue_golomb_long(gb); // difference_of_pic_nums_minus1
223 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
224 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
225 get_ue_golomb_31(gb);
233 * Parse NAL units of found picture and decode some basic information.
235 * @param s parser context.
236 * @param avctx codec context.
237 * @param buf buffer with field/frame data.
238 * @param buf_size size of the buffer.
240 static inline int parse_nal_units(AVCodecParserContext *s,
241 AVCodecContext *avctx,
242 const uint8_t * const buf, int buf_size)
244 H264ParseContext *p = s->priv_data;
245 H2645NAL nal = { NULL };
246 int buf_index, next_avc;
248 unsigned int slice_type;
249 int state = -1, got_reset = 0;
250 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
254 /* set some sane default values */
255 s->pict_type = AV_PICTURE_TYPE_I;
257 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
259 ff_h264_sei_uninit(&p->sei);
260 p->sei.frame_packing.frame_packing_arrangement_cancel_flag = -1;
266 next_avc = p->is_avc ? 0 : buf_size;
269 int src_length, consumed, nalsize = 0;
271 if (buf_index >= next_avc) {
272 nalsize = get_nalsize(p->nal_length_size, buf, buf_size, &buf_index, avctx);
275 next_avc = buf_index + nalsize;
277 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
278 if (buf_index >= buf_size)
280 if (buf_index >= next_avc)
283 src_length = next_avc - buf_index;
285 state = buf[buf_index];
286 switch (state & 0x1f) {
288 case H264_NAL_IDR_SLICE:
289 // Do not walk the whole buffer just to decode slice header
290 if ((state & 0x1f) == H264_NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
291 /* IDR or disposable slice
292 * No need to decode many bytes because MMCOs shall not be present. */
296 /* To decode up to MMCOs */
297 if (src_length > 1000)
302 consumed = ff_h2645_extract_rbsp(buf + buf_index, src_length, &nal, 1);
306 buf_index += consumed;
308 ret = init_get_bits8(&nal.gb, nal.data, nal.size);
312 nal.ref_idc = get_bits(&nal.gb, 2);
313 nal.type = get_bits(&nal.gb, 5);
317 ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps, 0);
320 ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
324 ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
326 case H264_NAL_IDR_SLICE:
329 p->poc.prev_frame_num = 0;
330 p->poc.prev_frame_num_offset = 0;
331 p->poc.prev_poc_msb =
332 p->poc.prev_poc_lsb = 0;
335 get_ue_golomb_long(&nal.gb); // skip first_mb_in_slice
336 slice_type = get_ue_golomb_31(&nal.gb);
337 s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
338 if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
339 /* key frame, since recovery_frame_cnt is set */
342 pps_id = get_ue_golomb(&nal.gb);
343 if (pps_id >= MAX_PPS_COUNT) {
344 av_log(avctx, AV_LOG_ERROR,
345 "pps_id %u out of range\n", pps_id);
348 if (!p->ps.pps_list[pps_id]) {
349 av_log(avctx, AV_LOG_ERROR,
350 "non-existing PPS %u referenced\n", pps_id);
354 av_buffer_unref(&p->ps.pps_ref);
355 av_buffer_unref(&p->ps.sps_ref);
358 p->ps.pps_ref = av_buffer_ref(p->ps.pps_list[pps_id]);
361 p->ps.pps = (const PPS*)p->ps.pps_ref->data;
363 if (!p->ps.sps_list[p->ps.pps->sps_id]) {
364 av_log(avctx, AV_LOG_ERROR,
365 "non-existing SPS %u referenced\n", p->ps.pps->sps_id);
369 p->ps.sps_ref = av_buffer_ref(p->ps.sps_list[p->ps.pps->sps_id]);
372 p->ps.sps = (const SPS*)p->ps.sps_ref->data;
376 // heuristic to detect non marked keyframes
377 if (p->ps.sps->ref_frame_count <= 1 && p->ps.pps->ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
380 p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
382 s->coded_width = 16 * sps->mb_width;
383 s->coded_height = 16 * sps->mb_height;
384 s->width = s->coded_width - (sps->crop_right + sps->crop_left);
385 s->height = s->coded_height - (sps->crop_top + sps->crop_bottom);
386 if (s->width <= 0 || s->height <= 0) {
387 s->width = s->coded_width;
388 s->height = s->coded_height;
391 switch (sps->bit_depth_luma) {
393 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P9;
394 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
395 else s->format = AV_PIX_FMT_YUV420P9;
398 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P10;
399 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
400 else s->format = AV_PIX_FMT_YUV420P10;
403 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P;
404 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
405 else s->format = AV_PIX_FMT_YUV420P;
408 s->format = AV_PIX_FMT_NONE;
411 avctx->profile = ff_h264_get_profile(sps);
412 avctx->level = sps->level_idc;
414 if (sps->frame_mbs_only_flag) {
415 p->picture_structure = PICT_FRAME;
417 if (get_bits1(&nal.gb)) { // field_pic_flag
418 p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
420 p->picture_structure = PICT_FRAME;
424 if (nal.type == H264_NAL_IDR_SLICE)
425 get_ue_golomb_long(&nal.gb); /* idr_pic_id */
426 if (sps->poc_type == 0) {
427 p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
429 if (p->ps.pps->pic_order_present == 1 &&
430 p->picture_structure == PICT_FRAME)
431 p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
434 if (sps->poc_type == 1 &&
435 !sps->delta_pic_order_always_zero_flag) {
436 p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
438 if (p->ps.pps->pic_order_present == 1 &&
439 p->picture_structure == PICT_FRAME)
440 p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
443 /* Decode POC of this picture.
444 * The prev_ values needed for decoding POC of the next picture are not set here. */
445 field_poc[0] = field_poc[1] = INT_MAX;
446 ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
447 &p->poc, p->picture_structure, nal.ref_idc);
449 /* Continue parsing to check if MMCO_RESET is present.
450 * FIXME: MMCO_RESET could appear in non-first slice.
451 * Maybe, we should parse all undisposable non-IDR slice of this
452 * picture until encountering MMCO_RESET in a slice of it. */
453 if (nal.ref_idc && nal.type != H264_NAL_IDR_SLICE) {
454 got_reset = scan_mmco_reset(s, &nal.gb, avctx);
459 /* Set up the prev_ values for decoding POC of the next picture. */
460 p->poc.prev_frame_num = got_reset ? 0 : p->poc.frame_num;
461 p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
462 if (nal.ref_idc != 0) {
464 p->poc.prev_poc_msb = p->poc.poc_msb;
465 p->poc.prev_poc_lsb = p->poc.poc_lsb;
467 p->poc.prev_poc_msb = 0;
468 p->poc.prev_poc_lsb =
469 p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
473 if (sps->pic_struct_present_flag) {
474 switch (p->sei.picture_timing.pic_struct) {
475 case SEI_PIC_STRUCT_TOP_FIELD:
476 case SEI_PIC_STRUCT_BOTTOM_FIELD:
479 case SEI_PIC_STRUCT_FRAME:
480 case SEI_PIC_STRUCT_TOP_BOTTOM:
481 case SEI_PIC_STRUCT_BOTTOM_TOP:
484 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
485 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
488 case SEI_PIC_STRUCT_FRAME_DOUBLING:
491 case SEI_PIC_STRUCT_FRAME_TRIPLING:
495 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
499 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
502 if (p->picture_structure == PICT_FRAME) {
503 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
504 if (sps->pic_struct_present_flag) {
505 switch (p->sei.picture_timing.pic_struct) {
506 case SEI_PIC_STRUCT_TOP_BOTTOM:
507 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
508 s->field_order = AV_FIELD_TT;
510 case SEI_PIC_STRUCT_BOTTOM_TOP:
511 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
512 s->field_order = AV_FIELD_BB;
515 s->field_order = AV_FIELD_PROGRESSIVE;
519 if (field_poc[0] < field_poc[1])
520 s->field_order = AV_FIELD_TT;
521 else if (field_poc[0] > field_poc[1])
522 s->field_order = AV_FIELD_BB;
524 s->field_order = AV_FIELD_PROGRESSIVE;
527 if (p->picture_structure == PICT_TOP_FIELD)
528 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
530 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
531 s->field_order = AV_FIELD_UNKNOWN;
534 av_freep(&nal.rbsp_buffer);
535 return 0; /* no need to evaluate the rest */
539 av_freep(&nal.rbsp_buffer);
542 /* didn't find a picture! */
543 av_log(avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
545 av_freep(&nal.rbsp_buffer);
549 static int h264_parse(AVCodecParserContext *s,
550 AVCodecContext *avctx,
551 const uint8_t **poutbuf, int *poutbuf_size,
552 const uint8_t *buf, int buf_size)
554 H264ParseContext *p = s->priv_data;
555 ParseContext *pc = &p->pc;
560 if (avctx->extradata_size) {
561 ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
562 &p->ps, &p->is_avc, &p->nal_length_size,
563 avctx->err_recognition, avctx);
567 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
570 next = h264_find_frame_end(p, buf, buf_size, avctx);
572 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
578 if (next < 0 && next != END_NOT_FOUND) {
579 av_assert1(pc->last_index + next >= 0);
580 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next, avctx); // update state
584 parse_nal_units(s, avctx, buf, buf_size);
586 if (avctx->framerate.num)
587 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
588 if (p->sei.picture_timing.cpb_removal_delay >= 0) {
589 s->dts_sync_point = p->sei.buffering_period.present;
590 s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
591 s->pts_dts_delta = p->sei.picture_timing.dpb_output_delay;
593 s->dts_sync_point = INT_MIN;
594 s->dts_ref_dts_delta = INT_MIN;
595 s->pts_dts_delta = INT_MIN;
598 if (s->flags & PARSER_FLAG_ONCE) {
599 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
602 if (s->dts_sync_point >= 0) {
603 int64_t den = avctx->time_base.den * (int64_t)avctx->pkt_timebase.num;
605 int64_t num = avctx->time_base.num * (int64_t)avctx->pkt_timebase.den;
606 if (s->dts != AV_NOPTS_VALUE) {
607 // got DTS from the stream, update reference timestamp
608 p->reference_dts = s->dts - av_rescale(s->dts_ref_dts_delta, num, den);
609 } else if (p->reference_dts != AV_NOPTS_VALUE) {
610 // compute DTS based on reference timestamp
611 s->dts = p->reference_dts + av_rescale(s->dts_ref_dts_delta, num, den);
614 if (p->reference_dts != AV_NOPTS_VALUE && s->pts == AV_NOPTS_VALUE)
615 s->pts = s->dts + av_rescale(s->pts_dts_delta, num, den);
617 if (s->dts_sync_point > 0)
618 p->reference_dts = s->dts; // new reference
623 *poutbuf_size = buf_size;
627 static int h264_split(AVCodecContext *avctx,
628 const uint8_t *buf, int buf_size)
633 const uint8_t *ptr = buf, *end = buf + buf_size;
637 ptr = avpriv_find_start_code(ptr, end, &state);
638 if ((state & 0xFFFFFF00) != 0x100)
640 nalu_type = state & 0x1F;
641 if (nalu_type == H264_NAL_SPS) {
643 } else if (nalu_type == H264_NAL_PPS)
645 /* else if (nalu_type == 0x01 ||
646 * nalu_type == 0x02 ||
647 * nalu_type == 0x05) {
650 else if ((nalu_type != H264_NAL_SEI || has_pps) &&
651 nalu_type != H264_NAL_AUD && nalu_type != H264_NAL_SPS_EXT &&
654 while (ptr - 4 > buf && ptr[-5] == 0)
656 return ptr - 4 - buf;
664 static void h264_close(AVCodecParserContext *s)
666 H264ParseContext *p = s->priv_data;
667 ParseContext *pc = &p->pc;
669 av_freep(&pc->buffer);
671 ff_h264_sei_uninit(&p->sei);
672 ff_h264_ps_uninit(&p->ps);
675 static av_cold int init(AVCodecParserContext *s)
677 H264ParseContext *p = s->priv_data;
679 p->reference_dts = AV_NOPTS_VALUE;
680 ff_h264dsp_init(&p->h264dsp, 8, 1);
684 AVCodecParser ff_h264_parser = {
685 .codec_ids = { AV_CODEC_ID_H264 },
686 .priv_data_size = sizeof(H264ParseContext),
688 .parser_parse = h264_parse,
689 .parser_close = h264_close,