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;
64 int last_frame_num, last_picture_structure;
68 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
69 int buf_size, void *logctx)
73 ParseContext *pc = &p->pc;
75 int next_avc = p->is_avc ? 0 : buf_size;
76 // mb_addr= pc->mb_addr - 1;
81 if (p->is_avc && !p->nal_length_size)
82 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
84 for (i = 0; i < buf_size; i++) {
88 for (j = 0; j < p->nal_length_size; j++)
89 nalsize = (nalsize << 8) | buf[i++];
90 if (nalsize <= 0 || nalsize > buf_size - i) {
91 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
94 next_avc = i + nalsize;
99 i += p->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
102 } else if (state <= 2) {
104 state ^= 5; // 2->7, 1->4, 0->5
108 state >>= 1; // 2->1, 1->0, 0->0
109 } else if (state <= 5) {
110 int nalu_type = buf[i] & 0x1F;
111 if (nalu_type == H264_NAL_SEI || nalu_type == H264_NAL_SPS ||
112 nalu_type == H264_NAL_PPS || nalu_type == H264_NAL_AUD) {
113 if (pc->frame_start_found) {
117 } else if (nalu_type == H264_NAL_SLICE || nalu_type == H264_NAL_DPA ||
118 nalu_type == H264_NAL_IDR_SLICE) {
124 unsigned int mb, last_mb = p->parse_last_mb;
126 p->parse_history[p->parse_history_count++] = buf[i];
128 init_get_bits(&gb, p->parse_history, 8*p->parse_history_count);
129 mb= get_ue_golomb_long(&gb);
130 if (get_bits_left(&gb) > 0 || p->parse_history_count > 5) {
131 p->parse_last_mb = mb;
132 if (pc->frame_start_found) {
134 i -= p->parse_history_count - 1;
135 p->parse_history_count = 0;
139 pc->frame_start_found = 1;
140 p->parse_history_count = 0;
148 return END_NOT_FOUND;
152 pc->frame_start_found = 0;
155 return i - (state & 5);
158 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
161 H264PredWeightTable pwt;
162 int slice_type_nos = s->pict_type & 3;
163 H264ParseContext *p = s->priv_data;
164 int list_count, ref_count[2];
167 if (p->ps.pps->redundant_pic_cnt_present)
168 get_ue_golomb(gb); // redundant_pic_count
170 if (slice_type_nos == AV_PICTURE_TYPE_B)
171 get_bits1(gb); // direct_spatial_mv_pred
173 if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
174 slice_type_nos, p->picture_structure, logctx) < 0)
175 return AVERROR_INVALIDDATA;
177 if (slice_type_nos != AV_PICTURE_TYPE_I) {
179 for (list = 0; list < list_count; list++) {
182 for (index = 0; ; index++) {
183 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
185 if (reordering_of_pic_nums_idc < 3)
186 get_ue_golomb_long(gb);
187 else if (reordering_of_pic_nums_idc > 3) {
188 av_log(logctx, AV_LOG_ERROR,
189 "illegal reordering_of_pic_nums_idc %d\n",
190 reordering_of_pic_nums_idc);
191 return AVERROR_INVALIDDATA;
195 if (index >= ref_count[list]) {
196 av_log(logctx, AV_LOG_ERROR,
197 "reference count %d overflow\n", index);
198 return AVERROR_INVALIDDATA;
205 if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
206 (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
207 ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
208 &pwt, p->picture_structure, logctx);
210 if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
212 for (i = 0; i < MAX_MMCO_COUNT; i++) {
213 MMCOOpcode opcode = get_ue_golomb_31(gb);
214 if (opcode > (unsigned) MMCO_LONG) {
215 av_log(logctx, AV_LOG_ERROR,
216 "illegal memory management control operation %d\n",
218 return AVERROR_INVALIDDATA;
220 if (opcode == MMCO_END)
222 else if (opcode == MMCO_RESET)
225 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
226 get_ue_golomb_long(gb); // difference_of_pic_nums_minus1
227 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
228 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
229 get_ue_golomb_31(gb);
237 * Parse NAL units of found picture and decode some basic information.
239 * @param s parser context.
240 * @param avctx codec context.
241 * @param buf buffer with field/frame data.
242 * @param buf_size size of the buffer.
244 static inline int parse_nal_units(AVCodecParserContext *s,
245 AVCodecContext *avctx,
246 const uint8_t * const buf, int buf_size)
248 H264ParseContext *p = s->priv_data;
249 H2645RBSP rbsp = { NULL };
250 H2645NAL nal = { NULL };
251 int buf_index, next_avc;
253 unsigned int slice_type;
254 int state = -1, got_reset = 0;
255 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
259 /* set some sane default values */
260 s->pict_type = AV_PICTURE_TYPE_I;
262 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
264 ff_h264_sei_uninit(&p->sei);
265 p->sei.frame_packing.arrangement_cancel_flag = -1;
270 av_fast_padded_malloc(&rbsp.rbsp_buffer, &rbsp.rbsp_buffer_alloc_size, buf_size);
271 if (!rbsp.rbsp_buffer)
272 return AVERROR(ENOMEM);
275 next_avc = p->is_avc ? 0 : buf_size;
278 int src_length, consumed, nalsize = 0;
280 if (buf_index >= next_avc) {
281 nalsize = get_nalsize(p->nal_length_size, buf, buf_size, &buf_index, avctx);
284 next_avc = buf_index + nalsize;
286 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
287 if (buf_index >= buf_size)
289 if (buf_index >= next_avc)
292 src_length = next_avc - buf_index;
294 state = buf[buf_index];
295 switch (state & 0x1f) {
297 case H264_NAL_IDR_SLICE:
298 // Do not walk the whole buffer just to decode slice header
299 if ((state & 0x1f) == H264_NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
300 /* IDR or disposable slice
301 * No need to decode many bytes because MMCOs shall not be present. */
305 /* To decode up to MMCOs */
306 if (src_length > 1000)
311 consumed = ff_h2645_extract_rbsp(buf + buf_index, src_length, &rbsp, &nal, 1);
315 buf_index += consumed;
317 ret = init_get_bits8(&nal.gb, nal.data, nal.size);
321 nal.ref_idc = get_bits(&nal.gb, 2);
322 nal.type = get_bits(&nal.gb, 5);
326 ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps, 0);
329 ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
333 ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
335 case H264_NAL_IDR_SLICE:
338 p->poc.prev_frame_num = 0;
339 p->poc.prev_frame_num_offset = 0;
340 p->poc.prev_poc_msb =
341 p->poc.prev_poc_lsb = 0;
344 get_ue_golomb_long(&nal.gb); // skip first_mb_in_slice
345 slice_type = get_ue_golomb_31(&nal.gb);
346 s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
347 if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
348 /* key frame, since recovery_frame_cnt is set */
351 pps_id = get_ue_golomb(&nal.gb);
352 if (pps_id >= MAX_PPS_COUNT) {
353 av_log(avctx, AV_LOG_ERROR,
354 "pps_id %u out of range\n", pps_id);
357 if (!p->ps.pps_list[pps_id]) {
358 av_log(avctx, AV_LOG_ERROR,
359 "non-existing PPS %u referenced\n", pps_id);
363 av_buffer_unref(&p->ps.pps_ref);
366 p->ps.pps_ref = av_buffer_ref(p->ps.pps_list[pps_id]);
369 p->ps.pps = (const PPS*)p->ps.pps_ref->data;
370 p->ps.sps = p->ps.pps->sps;
373 // heuristic to detect non marked keyframes
374 if (p->ps.sps->ref_frame_count <= 1 && p->ps.pps->ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
377 p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
379 s->coded_width = 16 * sps->mb_width;
380 s->coded_height = 16 * sps->mb_height;
381 s->width = s->coded_width - (sps->crop_right + sps->crop_left);
382 s->height = s->coded_height - (sps->crop_top + sps->crop_bottom);
383 if (s->width <= 0 || s->height <= 0) {
384 s->width = s->coded_width;
385 s->height = s->coded_height;
388 switch (sps->bit_depth_luma) {
390 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P9;
391 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
392 else s->format = AV_PIX_FMT_YUV420P9;
395 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P10;
396 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
397 else s->format = AV_PIX_FMT_YUV420P10;
400 if (sps->chroma_format_idc == 3) s->format = AV_PIX_FMT_YUV444P;
401 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
402 else s->format = AV_PIX_FMT_YUV420P;
405 s->format = AV_PIX_FMT_NONE;
408 avctx->profile = ff_h264_get_profile(sps);
409 avctx->level = sps->level_idc;
411 if (sps->frame_mbs_only_flag) {
412 p->picture_structure = PICT_FRAME;
414 if (get_bits1(&nal.gb)) { // field_pic_flag
415 p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
417 p->picture_structure = PICT_FRAME;
421 if (nal.type == H264_NAL_IDR_SLICE)
422 get_ue_golomb_long(&nal.gb); /* idr_pic_id */
423 if (sps->poc_type == 0) {
424 p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
426 if (p->ps.pps->pic_order_present == 1 &&
427 p->picture_structure == PICT_FRAME)
428 p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
431 if (sps->poc_type == 1 &&
432 !sps->delta_pic_order_always_zero_flag) {
433 p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
435 if (p->ps.pps->pic_order_present == 1 &&
436 p->picture_structure == PICT_FRAME)
437 p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
440 /* Decode POC of this picture.
441 * The prev_ values needed for decoding POC of the next picture are not set here. */
442 field_poc[0] = field_poc[1] = INT_MAX;
443 ret = ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
444 &p->poc, p->picture_structure, nal.ref_idc);
448 /* Continue parsing to check if MMCO_RESET is present.
449 * FIXME: MMCO_RESET could appear in non-first slice.
450 * Maybe, we should parse all undisposable non-IDR slice of this
451 * picture until encountering MMCO_RESET in a slice of it. */
452 if (nal.ref_idc && nal.type != H264_NAL_IDR_SLICE) {
453 got_reset = scan_mmco_reset(s, &nal.gb, avctx);
458 /* Set up the prev_ values for decoding POC of the next picture. */
459 p->poc.prev_frame_num = got_reset ? 0 : p->poc.frame_num;
460 p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
461 if (nal.ref_idc != 0) {
463 p->poc.prev_poc_msb = p->poc.poc_msb;
464 p->poc.prev_poc_lsb = p->poc.poc_lsb;
466 p->poc.prev_poc_msb = 0;
467 p->poc.prev_poc_lsb =
468 p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
472 if (p->sei.picture_timing.present) {
473 ret = ff_h264_sei_process_picture_timing(&p->sei.picture_timing,
476 av_log(avctx, AV_LOG_ERROR, "Error processing the picture timing SEI\n");
477 p->sei.picture_timing.present = 0;
481 if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
482 switch (p->sei.picture_timing.pic_struct) {
483 case H264_SEI_PIC_STRUCT_TOP_FIELD:
484 case H264_SEI_PIC_STRUCT_BOTTOM_FIELD:
487 case H264_SEI_PIC_STRUCT_FRAME:
488 case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
489 case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
492 case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
493 case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
496 case H264_SEI_PIC_STRUCT_FRAME_DOUBLING:
499 case H264_SEI_PIC_STRUCT_FRAME_TRIPLING:
503 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
507 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
510 if (p->picture_structure == PICT_FRAME) {
511 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
512 if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
513 switch (p->sei.picture_timing.pic_struct) {
514 case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
515 case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
516 s->field_order = AV_FIELD_TT;
518 case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
519 case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
520 s->field_order = AV_FIELD_BB;
523 s->field_order = AV_FIELD_PROGRESSIVE;
527 if (field_poc[0] < field_poc[1])
528 s->field_order = AV_FIELD_TT;
529 else if (field_poc[0] > field_poc[1])
530 s->field_order = AV_FIELD_BB;
532 s->field_order = AV_FIELD_PROGRESSIVE;
535 if (p->picture_structure == PICT_TOP_FIELD)
536 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
538 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
539 if (p->poc.frame_num == p->last_frame_num &&
540 p->last_picture_structure != AV_PICTURE_STRUCTURE_UNKNOWN &&
541 p->last_picture_structure != AV_PICTURE_STRUCTURE_FRAME &&
542 p->last_picture_structure != s->picture_structure) {
543 if (p->last_picture_structure == AV_PICTURE_STRUCTURE_TOP_FIELD)
544 s->field_order = AV_FIELD_TT;
546 s->field_order = AV_FIELD_BB;
548 s->field_order = AV_FIELD_UNKNOWN;
550 p->last_picture_structure = s->picture_structure;
551 p->last_frame_num = p->poc.frame_num;
554 av_freep(&rbsp.rbsp_buffer);
555 return 0; /* no need to evaluate the rest */
559 av_freep(&rbsp.rbsp_buffer);
562 /* didn't find a picture! */
563 av_log(avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
565 av_freep(&rbsp.rbsp_buffer);
569 static int h264_parse(AVCodecParserContext *s,
570 AVCodecContext *avctx,
571 const uint8_t **poutbuf, int *poutbuf_size,
572 const uint8_t *buf, int buf_size)
574 H264ParseContext *p = s->priv_data;
575 ParseContext *pc = &p->pc;
580 if (avctx->extradata_size) {
581 ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
582 &p->ps, &p->is_avc, &p->nal_length_size,
583 avctx->err_recognition, avctx);
587 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
590 next = h264_find_frame_end(p, buf, buf_size, avctx);
592 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
598 if (next < 0 && next != END_NOT_FOUND) {
599 av_assert1(pc->last_index + next >= 0);
600 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next, avctx); // update state
604 parse_nal_units(s, avctx, buf, buf_size);
606 if (avctx->framerate.num)
607 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
608 if (p->sei.picture_timing.cpb_removal_delay >= 0) {
609 s->dts_sync_point = p->sei.buffering_period.present;
610 s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
611 s->pts_dts_delta = p->sei.picture_timing.dpb_output_delay;
613 s->dts_sync_point = INT_MIN;
614 s->dts_ref_dts_delta = INT_MIN;
615 s->pts_dts_delta = INT_MIN;
618 if (s->flags & PARSER_FLAG_ONCE) {
619 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
622 if (s->dts_sync_point >= 0) {
623 int64_t den = avctx->time_base.den * (int64_t)avctx->pkt_timebase.num;
625 int64_t num = avctx->time_base.num * (int64_t)avctx->pkt_timebase.den;
626 if (s->dts != AV_NOPTS_VALUE) {
627 // got DTS from the stream, update reference timestamp
628 p->reference_dts = s->dts - av_rescale(s->dts_ref_dts_delta, num, den);
629 } else if (p->reference_dts != AV_NOPTS_VALUE) {
630 // compute DTS based on reference timestamp
631 s->dts = p->reference_dts + av_rescale(s->dts_ref_dts_delta, num, den);
634 if (p->reference_dts != AV_NOPTS_VALUE && s->pts == AV_NOPTS_VALUE)
635 s->pts = s->dts + av_rescale(s->pts_dts_delta, num, den);
637 if (s->dts_sync_point > 0)
638 p->reference_dts = s->dts; // new reference
643 *poutbuf_size = buf_size;
647 static void h264_close(AVCodecParserContext *s)
649 H264ParseContext *p = s->priv_data;
650 ParseContext *pc = &p->pc;
652 av_freep(&pc->buffer);
654 ff_h264_sei_uninit(&p->sei);
655 ff_h264_ps_uninit(&p->ps);
658 static av_cold int init(AVCodecParserContext *s)
660 H264ParseContext *p = s->priv_data;
662 p->reference_dts = AV_NOPTS_VALUE;
663 p->last_frame_num = INT_MAX;
664 ff_h264dsp_init(&p->h264dsp, 8, 1);
668 const AVCodecParser ff_h264_parser = {
669 .codec_ids = { AV_CODEC_ID_H264 },
670 .priv_data_size = sizeof(H264ParseContext),
672 .parser_parse = h264_parse,
673 .parser_close = h264_close,