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 / MPEG4 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"
44 #include "mpegutils.h"
47 typedef struct H264ParseContext {
54 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
57 H264Context *h = &p->h;
60 ParseContext *pc = &p->pc;
62 int next_avc= h->is_avc ? 0 : buf_size;
63 // mb_addr= pc->mb_addr - 1;
68 if (h->is_avc && !h->nal_length_size)
69 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
71 for (i = 0; i < buf_size; i++) {
75 for (j = 0; j < h->nal_length_size; j++)
76 nalsize = (nalsize << 8) | buf[i++];
77 if (nalsize <= 0 || nalsize > buf_size - i) {
78 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
81 next_avc = i + nalsize;
86 i += h->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
89 } else if (state <= 2) {
91 state ^= 5; // 2->7, 1->4, 0->5
95 state >>= 1; // 2->1, 1->0, 0->0
96 } else if (state <= 5) {
97 int nalu_type = buf[i] & 0x1F;
98 if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
99 nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
100 if (pc->frame_start_found) {
104 } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
105 nalu_type == NAL_IDR_SLICE) {
111 h->parse_history[h->parse_history_count++]= buf[i];
112 if (h->parse_history_count>5) {
113 unsigned int mb, last_mb= h->parse_last_mb;
116 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
117 h->parse_history_count=0;
118 mb= get_ue_golomb_long(&gb);
119 h->parse_last_mb= mb;
120 if (pc->frame_start_found) {
124 pc->frame_start_found = 1;
132 return END_NOT_FOUND;
136 pc->frame_start_found = 0;
139 return i - (state & 5) - 5 * (state > 7);
142 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb)
144 H264PredWeightTable pwt;
145 int slice_type_nos = s->pict_type & 3;
146 H264ParseContext *p = s->priv_data;
147 H264Context *h = &p->h;
148 int list_count, ref_count[2];
151 if (h->pps.redundant_pic_cnt_present)
152 get_ue_golomb(gb); // redundant_pic_count
154 if (slice_type_nos == AV_PICTURE_TYPE_B)
155 get_bits1(gb); // direct_spatial_mv_pred
157 if (ff_h264_parse_ref_count(&list_count, ref_count, gb, &h->pps,
158 slice_type_nos, h->picture_structure, h->avctx) < 0)
159 return AVERROR_INVALIDDATA;
161 if (slice_type_nos != AV_PICTURE_TYPE_I) {
163 for (list = 0; list < list_count; list++) {
166 for (index = 0; ; index++) {
167 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
169 if (reordering_of_pic_nums_idc < 3)
170 get_ue_golomb_long(gb);
171 else if (reordering_of_pic_nums_idc > 3) {
172 av_log(h->avctx, AV_LOG_ERROR,
173 "illegal reordering_of_pic_nums_idc %d\n",
174 reordering_of_pic_nums_idc);
175 return AVERROR_INVALIDDATA;
179 if (index >= ref_count[list]) {
180 av_log(h->avctx, AV_LOG_ERROR,
181 "reference count %d overflow\n", index);
182 return AVERROR_INVALIDDATA;
189 if ((h->pps.weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
190 (h->pps.weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
191 ff_h264_pred_weight_table(gb, &h->sps, ref_count, slice_type_nos,
194 if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
196 for (i = 0; i < MAX_MMCO_COUNT; i++) {
197 MMCOOpcode opcode = get_ue_golomb_31(gb);
198 if (opcode > (unsigned) MMCO_LONG) {
199 av_log(h->avctx, AV_LOG_ERROR,
200 "illegal memory management control operation %d\n",
202 return AVERROR_INVALIDDATA;
204 if (opcode == MMCO_END)
206 else if (opcode == MMCO_RESET)
209 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
210 get_ue_golomb_long(gb); // difference_of_pic_nums_minus1
211 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
212 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
213 get_ue_golomb_31(gb);
221 * Parse NAL units of found picture and decode some basic information.
223 * @param s parser context.
224 * @param avctx codec context.
225 * @param buf buffer with field/frame data.
226 * @param buf_size size of the buffer.
228 static inline int parse_nal_units(AVCodecParserContext *s,
229 AVCodecContext *avctx,
230 const uint8_t * const buf, int buf_size)
232 H264ParseContext *p = s->priv_data;
233 H264Context *h = &p->h;
234 H2645NAL nal = { NULL };
235 int buf_index, next_avc;
237 unsigned int slice_type;
238 int state = -1, got_reset = 0;
239 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
243 /* set some sane default values */
244 s->pict_type = AV_PICTURE_TYPE_I;
246 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
249 ff_h264_reset_sei(h);
250 h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
256 next_avc = h->is_avc ? 0 : buf_size;
258 int src_length, consumed, nalsize = 0;
260 if (buf_index >= next_avc) {
261 nalsize = get_avc_nalsize(h, buf, buf_size, &buf_index);
264 next_avc = buf_index + nalsize;
266 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
267 if (buf_index >= buf_size)
269 if (buf_index >= next_avc)
272 src_length = next_avc - buf_index;
274 state = buf[buf_index];
275 switch (state & 0x1f) {
278 // Do not walk the whole buffer just to decode slice header
279 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
280 /* IDR or disposable slice
281 * No need to decode many bytes because MMCOs shall not be present. */
285 /* To decode up to MMCOs */
286 if (src_length > 1000)
291 consumed = ff_h2645_extract_rbsp(buf + buf_index, src_length, &nal);
295 buf_index += consumed;
297 ret = init_get_bits8(&nal.gb, nal.data, nal.size);
301 nal.ref_idc = get_bits(&nal.gb, 2);
302 nal.type = get_bits(&nal.gb, 5);
305 h->nal_ref_idc = nal.ref_idc;
306 h->nal_unit_type = nal.type;
308 switch (h->nal_unit_type) {
310 ff_h264_decode_seq_parameter_set(h, 0);
313 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
316 ff_h264_decode_sei(h);
321 h->prev_frame_num = 0;
322 h->prev_frame_num_offset = 0;
327 get_ue_golomb_long(&nal.gb); // skip first_mb_in_slice
328 slice_type = get_ue_golomb_31(&nal.gb);
329 s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
330 if (h->sei_recovery_frame_cnt >= 0) {
331 /* key frame, since recovery_frame_cnt is set */
334 pps_id = get_ue_golomb(&nal.gb);
335 if (pps_id >= MAX_PPS_COUNT) {
336 av_log(h->avctx, AV_LOG_ERROR,
337 "pps_id %u out of range\n", pps_id);
340 if (!h->pps_buffers[pps_id]) {
341 av_log(h->avctx, AV_LOG_ERROR,
342 "non-existing PPS %u referenced\n", pps_id);
345 h->pps = *h->pps_buffers[pps_id];
346 if (!h->sps_buffers[h->pps.sps_id]) {
347 av_log(h->avctx, AV_LOG_ERROR,
348 "non-existing SPS %u referenced\n", h->pps.sps_id);
351 h->sps = *h->sps_buffers[h->pps.sps_id];
352 h->frame_num = get_bits(&nal.gb, h->sps.log2_max_frame_num);
354 if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
357 s->coded_width = 16 * h->sps.mb_width;
358 s->coded_height = 16 * h->sps.mb_height;
359 s->width = s->coded_width - (h->sps.crop_right + h->sps.crop_left);
360 s->height = s->coded_height - (h->sps.crop_top + h->sps.crop_bottom);
361 if (s->width <= 0 || s->height <= 0) {
362 s->width = s->coded_width;
363 s->height = s->coded_height;
366 switch (h->sps.bit_depth_luma) {
368 if (CHROMA444(h)) s->format = AV_PIX_FMT_YUV444P9;
369 else if (CHROMA422(h)) s->format = AV_PIX_FMT_YUV422P9;
370 else s->format = AV_PIX_FMT_YUV420P9;
373 if (CHROMA444(h)) s->format = AV_PIX_FMT_YUV444P10;
374 else if (CHROMA422(h)) s->format = AV_PIX_FMT_YUV422P10;
375 else s->format = AV_PIX_FMT_YUV420P10;
378 if (CHROMA444(h)) s->format = AV_PIX_FMT_YUV444P;
379 else if (CHROMA422(h)) s->format = AV_PIX_FMT_YUV422P;
380 else s->format = AV_PIX_FMT_YUV420P;
383 s->format = AV_PIX_FMT_NONE;
386 avctx->profile = ff_h264_get_profile(&h->sps);
387 avctx->level = h->sps.level_idc;
389 if (h->sps.frame_mbs_only_flag) {
390 h->picture_structure = PICT_FRAME;
392 if (get_bits1(&nal.gb)) { // field_pic_flag
393 h->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
395 h->picture_structure = PICT_FRAME;
399 if (h->nal_unit_type == NAL_IDR_SLICE)
400 get_ue_golomb_long(&nal.gb); /* idr_pic_id */
401 if (h->sps.poc_type == 0) {
402 h->poc_lsb = get_bits(&nal.gb, h->sps.log2_max_poc_lsb);
404 if (h->pps.pic_order_present == 1 &&
405 h->picture_structure == PICT_FRAME)
406 h->delta_poc_bottom = get_se_golomb(&nal.gb);
409 if (h->sps.poc_type == 1 &&
410 !h->sps.delta_pic_order_always_zero_flag) {
411 h->delta_poc[0] = get_se_golomb(&nal.gb);
413 if (h->pps.pic_order_present == 1 &&
414 h->picture_structure == PICT_FRAME)
415 h->delta_poc[1] = get_se_golomb(&nal.gb);
418 /* Decode POC of this picture.
419 * The prev_ values needed for decoding POC of the next picture are not set here. */
420 field_poc[0] = field_poc[1] = INT_MAX;
421 ff_init_poc(h, field_poc, &s->output_picture_number);
423 /* Continue parsing to check if MMCO_RESET is present.
424 * FIXME: MMCO_RESET could appear in non-first slice.
425 * Maybe, we should parse all undisposable non-IDR slice of this
426 * picture until encountering MMCO_RESET in a slice of it. */
427 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
428 got_reset = scan_mmco_reset(s, &nal.gb);
433 /* Set up the prev_ values for decoding POC of the next picture. */
434 h->prev_frame_num = got_reset ? 0 : h->frame_num;
435 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
436 if (h->nal_ref_idc != 0) {
438 h->prev_poc_msb = h->poc_msb;
439 h->prev_poc_lsb = h->poc_lsb;
443 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
447 if (h->sps.pic_struct_present_flag) {
448 switch (h->sei_pic_struct) {
449 case SEI_PIC_STRUCT_TOP_FIELD:
450 case SEI_PIC_STRUCT_BOTTOM_FIELD:
453 case SEI_PIC_STRUCT_FRAME:
454 case SEI_PIC_STRUCT_TOP_BOTTOM:
455 case SEI_PIC_STRUCT_BOTTOM_TOP:
458 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
459 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
462 case SEI_PIC_STRUCT_FRAME_DOUBLING:
465 case SEI_PIC_STRUCT_FRAME_TRIPLING:
469 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
473 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
476 if (h->picture_structure == PICT_FRAME) {
477 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
478 if (h->sps.pic_struct_present_flag) {
479 switch (h->sei_pic_struct) {
480 case SEI_PIC_STRUCT_TOP_BOTTOM:
481 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
482 s->field_order = AV_FIELD_TT;
484 case SEI_PIC_STRUCT_BOTTOM_TOP:
485 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
486 s->field_order = AV_FIELD_BB;
489 s->field_order = AV_FIELD_PROGRESSIVE;
493 if (field_poc[0] < field_poc[1])
494 s->field_order = AV_FIELD_TT;
495 else if (field_poc[0] > field_poc[1])
496 s->field_order = AV_FIELD_BB;
498 s->field_order = AV_FIELD_PROGRESSIVE;
501 if (h->picture_structure == PICT_TOP_FIELD)
502 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
504 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
505 s->field_order = AV_FIELD_UNKNOWN;
508 av_freep(&nal.rbsp_buffer);
509 return 0; /* no need to evaluate the rest */
513 av_freep(&nal.rbsp_buffer);
516 /* didn't find a picture! */
517 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
519 av_freep(&nal.rbsp_buffer);
523 static int h264_parse(AVCodecParserContext *s,
524 AVCodecContext *avctx,
525 const uint8_t **poutbuf, int *poutbuf_size,
526 const uint8_t *buf, int buf_size)
528 H264ParseContext *p = s->priv_data;
529 H264Context *h = &p->h;
530 ParseContext *pc = &p->pc;
535 if (avctx->extradata_size) {
537 // must be done like in decoder, otherwise opening the parser,
538 // letting it create extradata and then closing and opening again
539 // will cause has_b_frames to be always set.
540 // Note that estimate_timings_from_pts does exactly this.
541 if (!avctx->has_b_frames)
543 ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
547 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
550 next = h264_find_frame_end(p, buf, buf_size);
552 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
558 if (next < 0 && next != END_NOT_FOUND) {
559 av_assert1(pc->last_index + next >= 0);
560 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next); // update state
564 parse_nal_units(s, avctx, buf, buf_size);
566 if (avctx->framerate.num)
567 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
568 if (h->sei_cpb_removal_delay >= 0) {
569 s->dts_sync_point = h->sei_buffering_period_present;
570 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
571 s->pts_dts_delta = h->sei_dpb_output_delay;
573 s->dts_sync_point = INT_MIN;
574 s->dts_ref_dts_delta = INT_MIN;
575 s->pts_dts_delta = INT_MIN;
578 if (s->flags & PARSER_FLAG_ONCE) {
579 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
583 *poutbuf_size = buf_size;
587 static int h264_split(AVCodecContext *avctx,
588 const uint8_t *buf, int buf_size)
593 const uint8_t *ptr = buf, *end = buf + buf_size;
597 ptr = avpriv_find_start_code(ptr, end, &state);
598 if ((state & 0xFFFFFF00) != 0x100)
600 nalu_type = state & 0x1F;
601 if (nalu_type == NAL_SPS) {
603 } else if (nalu_type == NAL_PPS)
605 /* else if (nalu_type == 0x01 ||
606 * nalu_type == 0x02 ||
607 * nalu_type == 0x05) {
610 else if ((nalu_type != NAL_SEI || has_pps) &&
611 nalu_type != NAL_AUD && nalu_type != NAL_SPS_EXT &&
614 while (ptr - 4 > buf && ptr[-5] == 0)
616 return ptr - 4 - buf;
624 static void h264_close(AVCodecParserContext *s)
626 H264ParseContext *p = s->priv_data;
627 H264Context *h = &p->h;
628 ParseContext *pc = &p->pc;
630 av_freep(&pc->buffer);
631 ff_h264_free_context(h);
634 static av_cold int init(AVCodecParserContext *s)
636 H264ParseContext *p = s->priv_data;
637 H264Context *h = &p->h;
639 h->slice_ctx = av_mallocz(sizeof(*h->slice_ctx));
644 h->slice_context_count = 1;
645 ff_h264dsp_init(&h->h264dsp, 8, 1);
649 AVCodecParser ff_h264_parser = {
650 .codec_ids = { AV_CODEC_ID_H264 },
651 .priv_data_size = sizeof(H264ParseContext),
653 .parser_parse = h264_parse,
654 .parser_close = h264_close,