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
30 #include "libavutil/attributes.h"
35 #include "mpegutils.h"
37 typedef struct H264ParseContext {
44 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
47 H264Context *h = &p->h;
50 ParseContext *pc = &p->pc;
52 int next_avc= h->is_avc ? 0 : buf_size;
53 // mb_addr= pc->mb_addr - 1;
58 if (h->is_avc && !h->nal_length_size)
59 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
61 for (i = 0; i < buf_size; i++) {
65 for (j = 0; j < h->nal_length_size; j++)
66 nalsize = (nalsize << 8) | buf[i++];
67 if (nalsize <= 0 || nalsize > buf_size - i) {
68 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
71 next_avc = i + nalsize;
76 i += h->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
79 } else if (state <= 2) {
81 state ^= 5; // 2->7, 1->4, 0->5
85 state >>= 1; // 2->1, 1->0, 0->0
86 } else if (state <= 5) {
87 int nalu_type = buf[i] & 0x1F;
88 if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
89 nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
90 if (pc->frame_start_found) {
94 } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
95 nalu_type == NAL_IDR_SLICE) {
101 h->parse_history[h->parse_history_count++]= buf[i];
102 if (h->parse_history_count>5) {
103 unsigned int mb, last_mb= h->parse_last_mb;
106 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
107 h->parse_history_count=0;
108 mb= get_ue_golomb_long(&gb);
109 h->parse_last_mb= mb;
110 if (pc->frame_start_found) {
114 pc->frame_start_found = 1;
122 return END_NOT_FOUND;
126 pc->frame_start_found = 0;
129 return i - (state & 5) - 5 * (state > 7);
132 static int scan_mmco_reset(AVCodecParserContext *s)
134 H264ParseContext *p = s->priv_data;
135 H264Context *h = &p->h;
137 h->slice_type_nos = s->pict_type & 3;
139 if (h->pps.redundant_pic_cnt_present)
140 get_ue_golomb(&h->gb); // redundant_pic_count
142 if (ff_set_ref_count(h) < 0)
143 return AVERROR_INVALIDDATA;
145 if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
147 for (list = 0; list < h->list_count; list++) {
148 if (get_bits1(&h->gb)) {
150 for (index = 0; ; index++) {
151 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
153 if (reordering_of_pic_nums_idc < 3)
154 get_ue_golomb(&h->gb);
155 else if (reordering_of_pic_nums_idc > 3) {
156 av_log(h->avctx, AV_LOG_ERROR,
157 "illegal reordering_of_pic_nums_idc %d\n",
158 reordering_of_pic_nums_idc);
159 return AVERROR_INVALIDDATA;
163 if (index >= h->ref_count[list]) {
164 av_log(h->avctx, AV_LOG_ERROR,
165 "reference count %d overflow\n", index);
166 return AVERROR_INVALIDDATA;
173 if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) ||
174 (h->pps.weighted_bipred_idc == 1 && h->slice_type_nos == AV_PICTURE_TYPE_B))
175 ff_pred_weight_table(h);
177 if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
179 for (i = 0; i < MAX_MMCO_COUNT; i++) {
180 MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
181 if (opcode > (unsigned) MMCO_LONG) {
182 av_log(h->avctx, AV_LOG_ERROR,
183 "illegal memory management control operation %d\n",
185 return AVERROR_INVALIDDATA;
187 if (opcode == MMCO_END)
189 else if (opcode == MMCO_RESET)
192 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
193 get_ue_golomb(&h->gb);
194 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
195 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
196 get_ue_golomb_31(&h->gb);
204 * Parse NAL units of found picture and decode some basic information.
206 * @param s parser context.
207 * @param avctx codec context.
208 * @param buf buffer with field/frame data.
209 * @param buf_size size of the buffer.
211 static inline int parse_nal_units(AVCodecParserContext *s,
212 AVCodecContext *avctx,
213 const uint8_t * const buf, int buf_size)
215 H264ParseContext *p = s->priv_data;
216 H264Context *h = &p->h;
217 int buf_index, next_avc;
219 unsigned int slice_type;
220 int state = -1, got_reset = 0;
222 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
225 /* set some sane default values */
226 s->pict_type = AV_PICTURE_TYPE_I;
228 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
231 ff_h264_reset_sei(h);
232 h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
238 next_avc = h->is_avc ? 0 : buf_size;
240 int src_length, dst_length, consumed, nalsize = 0;
242 if (buf_index >= next_avc) {
243 nalsize = get_avc_nalsize(h, buf, buf_size, &buf_index);
246 next_avc = buf_index + nalsize;
248 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
249 if (buf_index >= buf_size)
251 if (buf_index >= next_avc)
254 src_length = next_avc - buf_index;
256 state = buf[buf_index];
257 switch (state & 0x1f) {
260 // Do not walk the whole buffer just to decode slice header
261 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
262 /* IDR or disposable slice
263 * No need to decode many bytes because MMCOs shall not be present. */
267 /* To decode up to MMCOs */
268 if (src_length > 1000)
273 ptr = ff_h264_decode_nal(h, buf + buf_index, &dst_length,
274 &consumed, src_length);
275 if (!ptr || dst_length < 0)
278 buf_index += consumed;
280 init_get_bits(&h->gb, ptr, 8 * dst_length);
281 switch (h->nal_unit_type) {
283 ff_h264_decode_seq_parameter_set(h);
286 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
289 ff_h264_decode_sei(h);
294 h->prev_frame_num = 0;
295 h->prev_frame_num_offset = 0;
300 get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
301 slice_type = get_ue_golomb_31(&h->gb);
302 s->pict_type = golomb_to_pict_type[slice_type % 5];
303 if (h->sei_recovery_frame_cnt >= 0) {
304 /* key frame, since recovery_frame_cnt is set */
307 pps_id = get_ue_golomb(&h->gb);
308 if (pps_id >= MAX_PPS_COUNT) {
309 av_log(h->avctx, AV_LOG_ERROR,
310 "pps_id %u out of range\n", pps_id);
313 if (!h->pps_buffers[pps_id]) {
314 av_log(h->avctx, AV_LOG_ERROR,
315 "non-existing PPS %u referenced\n", pps_id);
318 h->pps = *h->pps_buffers[pps_id];
319 if (!h->sps_buffers[h->pps.sps_id]) {
320 av_log(h->avctx, AV_LOG_ERROR,
321 "non-existing SPS %u referenced\n", h->pps.sps_id);
324 h->sps = *h->sps_buffers[h->pps.sps_id];
325 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
327 if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
330 avctx->profile = ff_h264_get_profile(&h->sps);
331 avctx->level = h->sps.level_idc;
333 if (h->sps.frame_mbs_only_flag) {
334 h->picture_structure = PICT_FRAME;
336 if (get_bits1(&h->gb)) { // field_pic_flag
337 h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
339 h->picture_structure = PICT_FRAME;
343 if (h->nal_unit_type == NAL_IDR_SLICE)
344 get_ue_golomb(&h->gb); /* idr_pic_id */
345 if (h->sps.poc_type == 0) {
346 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
348 if (h->pps.pic_order_present == 1 &&
349 h->picture_structure == PICT_FRAME)
350 h->delta_poc_bottom = get_se_golomb(&h->gb);
353 if (h->sps.poc_type == 1 &&
354 !h->sps.delta_pic_order_always_zero_flag) {
355 h->delta_poc[0] = get_se_golomb(&h->gb);
357 if (h->pps.pic_order_present == 1 &&
358 h->picture_structure == PICT_FRAME)
359 h->delta_poc[1] = get_se_golomb(&h->gb);
362 /* Decode POC of this picture.
363 * The prev_ values needed for decoding POC of the next picture are not set here. */
364 field_poc[0] = field_poc[1] = INT_MAX;
365 ff_init_poc(h, field_poc, &s->output_picture_number);
367 /* Continue parsing to check if MMCO_RESET is present.
368 * FIXME: MMCO_RESET could appear in non-first slice.
369 * Maybe, we should parse all undisposable non-IDR slice of this
370 * picture until encountering MMCO_RESET in a slice of it. */
371 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
372 got_reset = scan_mmco_reset(s);
377 /* Set up the prev_ values for decoding POC of the next picture. */
378 h->prev_frame_num = got_reset ? 0 : h->frame_num;
379 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
380 if (h->nal_ref_idc != 0) {
382 h->prev_poc_msb = h->poc_msb;
383 h->prev_poc_lsb = h->poc_lsb;
387 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
391 if (h->sps.pic_struct_present_flag) {
392 switch (h->sei_pic_struct) {
393 case SEI_PIC_STRUCT_TOP_FIELD:
394 case SEI_PIC_STRUCT_BOTTOM_FIELD:
397 case SEI_PIC_STRUCT_FRAME:
398 case SEI_PIC_STRUCT_TOP_BOTTOM:
399 case SEI_PIC_STRUCT_BOTTOM_TOP:
402 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
403 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
406 case SEI_PIC_STRUCT_FRAME_DOUBLING:
409 case SEI_PIC_STRUCT_FRAME_TRIPLING:
413 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
417 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
420 if (h->picture_structure == PICT_FRAME) {
421 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
422 if (h->sps.pic_struct_present_flag) {
423 switch (h->sei_pic_struct) {
424 case SEI_PIC_STRUCT_TOP_BOTTOM:
425 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
426 s->field_order = AV_FIELD_TT;
428 case SEI_PIC_STRUCT_BOTTOM_TOP:
429 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
430 s->field_order = AV_FIELD_BB;
433 s->field_order = AV_FIELD_PROGRESSIVE;
437 if (field_poc[0] < field_poc[1])
438 s->field_order = AV_FIELD_TT;
439 else if (field_poc[0] > field_poc[1])
440 s->field_order = AV_FIELD_BB;
442 s->field_order = AV_FIELD_PROGRESSIVE;
445 if (h->picture_structure == PICT_TOP_FIELD)
446 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
448 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
449 s->field_order = AV_FIELD_UNKNOWN;
452 return 0; /* no need to evaluate the rest */
457 /* didn't find a picture! */
458 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
462 static int h264_parse(AVCodecParserContext *s,
463 AVCodecContext *avctx,
464 const uint8_t **poutbuf, int *poutbuf_size,
465 const uint8_t *buf, int buf_size)
467 H264ParseContext *p = s->priv_data;
468 H264Context *h = &p->h;
469 ParseContext *pc = &p->pc;
474 if (avctx->extradata_size) {
476 // must be done like in decoder, otherwise opening the parser,
477 // letting it create extradata and then closing and opening again
478 // will cause has_b_frames to be always set.
479 // Note that estimate_timings_from_pts does exactly this.
480 if (!avctx->has_b_frames)
482 ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
486 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
489 next = h264_find_frame_end(p, buf, buf_size);
491 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
497 if (next < 0 && next != END_NOT_FOUND) {
498 av_assert1(pc->last_index + next >= 0);
499 h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next); // update state
503 parse_nal_units(s, avctx, buf, buf_size);
505 if (avctx->framerate.num)
506 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
507 if (h->sei_cpb_removal_delay >= 0) {
508 s->dts_sync_point = h->sei_buffering_period_present;
509 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
510 s->pts_dts_delta = h->sei_dpb_output_delay;
512 s->dts_sync_point = INT_MIN;
513 s->dts_ref_dts_delta = INT_MIN;
514 s->pts_dts_delta = INT_MIN;
517 if (s->flags & PARSER_FLAG_ONCE) {
518 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
522 *poutbuf_size = buf_size;
526 static int h264_split(AVCodecContext *avctx,
527 const uint8_t *buf, int buf_size)
534 for (i = 0; i <= buf_size; i++) {
535 if ((state & 0xFFFFFF1F) == 0x107)
537 if ((state & 0xFFFFFF1F) == 0x108)
539 /* if ((state&0xFFFFFF1F) == 0x101 ||
540 * (state&0xFFFFFF1F) == 0x102 ||
541 * (state&0xFFFFFF1F) == 0x105) {
544 if ((state & 0xFFFFFF00) == 0x100 && ((state & 0xFFFFFF1F) != 0x106 || has_pps) &&
545 (state & 0xFFFFFF1F) != 0x107 && (state & 0xFFFFFF1F) != 0x108 &&
546 (state & 0xFFFFFF1F) != 0x109 && (state & 0xFFFFFF1F) != 0x10d &&
547 (state & 0xFFFFFF1F) != 0x10f) {
549 while (i > 4 && buf[i - 5] == 0)
555 state = (state << 8) | buf[i];
560 static void h264_close(AVCodecParserContext *s)
562 H264ParseContext *p = s->priv_data;
563 H264Context *h = &p->h;
564 ParseContext *pc = &p->pc;
566 av_freep(&pc->buffer);
567 ff_h264_free_context(h);
570 static av_cold int init(AVCodecParserContext *s)
572 H264ParseContext *p = s->priv_data;
573 H264Context *h = &p->h;
574 h->thread_context[0] = h;
575 h->slice_context_count = 1;
576 ff_h264dsp_init(&h->h264dsp, 8, 1);
580 AVCodecParser ff_h264_parser = {
581 .codec_ids = { AV_CODEC_ID_H264 },
582 .priv_data_size = sizeof(H264ParseContext),
584 .parser_parse = h264_parse,
585 .parser_close = h264_close,