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"
38 static int h264_find_frame_end(H264Context *h, const uint8_t *buf,
43 ParseContext *pc = &h->parse_context;
44 int next_avc= h->is_avc ? 0 : buf_size;
46 // mb_addr= pc->mb_addr - 1;
51 if (h->is_avc && !h->nal_length_size)
52 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
54 for (i = 0; i < buf_size; i++) {
58 for (j = 0; j < h->nal_length_size; j++)
59 nalsize = (nalsize << 8) | buf[i++];
60 if (nalsize <= 0 || nalsize > buf_size - i) {
61 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
64 next_avc = i + nalsize;
69 i += h->h264dsp.h264_find_start_code_candidate(buf + i, next_avc - i);
72 } else if (state <= 2) {
74 state ^= 5; // 2->7, 1->4, 0->5
78 state >>= 1; // 2->1, 1->0, 0->0
79 } else if (state <= 5) {
80 int nalu_type = buf[i] & 0x1F;
81 if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
82 nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
83 if (pc->frame_start_found) {
87 } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
88 nalu_type == NAL_IDR_SLICE) {
94 h->parse_history[h->parse_history_count++]= buf[i];
95 if (h->parse_history_count>3) {
96 unsigned int mb, last_mb= h->parse_last_mb;
99 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
100 h->parse_history_count=0;
101 mb= get_ue_golomb_long(&gb);
102 last_mb= h->parse_last_mb;
103 h->parse_last_mb= mb;
104 if (pc->frame_start_found) {
108 pc->frame_start_found = 1;
116 return END_NOT_FOUND;
120 pc->frame_start_found = 0;
123 return i - (state & 5) - 3 * (state > 7);
126 static int scan_mmco_reset(AVCodecParserContext *s)
128 H264Context *h = s->priv_data;
130 h->slice_type_nos = s->pict_type & 3;
132 if (h->pps.redundant_pic_cnt_present)
133 get_ue_golomb(&h->gb); // redundant_pic_count
135 if (ff_set_ref_count(h) < 0)
136 return AVERROR_INVALIDDATA;
138 if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
140 for (list = 0; list < h->list_count; list++) {
141 if (get_bits1(&h->gb)) {
143 for (index = 0; ; index++) {
144 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
146 if (reordering_of_pic_nums_idc < 3)
147 get_ue_golomb(&h->gb);
148 else if (reordering_of_pic_nums_idc > 3) {
149 av_log(h->avctx, AV_LOG_ERROR,
150 "illegal reordering_of_pic_nums_idc %d\n",
151 reordering_of_pic_nums_idc);
152 return AVERROR_INVALIDDATA;
156 if (index >= h->ref_count[list]) {
157 av_log(h->avctx, AV_LOG_ERROR,
158 "reference count %d overflow\n", index);
159 return AVERROR_INVALIDDATA;
166 if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) ||
167 (h->pps.weighted_bipred_idc == 1 && h->slice_type_nos == AV_PICTURE_TYPE_B))
168 ff_pred_weight_table(h);
170 if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
172 for (i = 0; i < MAX_MMCO_COUNT; i++) {
173 MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
174 if (opcode > (unsigned) MMCO_LONG) {
175 av_log(h->avctx, AV_LOG_ERROR,
176 "illegal memory management control operation %d\n",
178 return AVERROR_INVALIDDATA;
180 if (opcode == MMCO_END)
182 else if (opcode == MMCO_RESET)
185 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
186 get_ue_golomb(&h->gb);
187 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
188 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
189 get_ue_golomb_31(&h->gb);
197 * Parse NAL units of found picture and decode some basic information.
199 * @param s parser context.
200 * @param avctx codec context.
201 * @param buf buffer with field/frame data.
202 * @param buf_size size of the buffer.
204 static inline int parse_nal_units(AVCodecParserContext *s,
205 AVCodecContext *avctx,
206 const uint8_t *buf, int buf_size)
208 H264Context *h = s->priv_data;
209 const uint8_t *buf_end = buf + buf_size;
211 unsigned int slice_type;
212 int state = -1, got_reset = 0;
214 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
217 /* set some sane default values */
218 s->pict_type = AV_PICTURE_TYPE_I;
220 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
223 ff_h264_reset_sei(h);
224 h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
230 int src_length, dst_length, consumed, nalsize = 0;
233 if (h->nal_length_size >= buf_end - buf) break;
235 for (i = 0; i < h->nal_length_size; i++)
236 nalsize = (nalsize << 8) | *buf++;
237 if (nalsize <= 0 || nalsize > buf_end - buf) {
238 av_log(h->avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
241 src_length = nalsize;
243 buf = avpriv_find_start_code(buf, buf_end, &state);
247 src_length = buf_end - buf;
249 switch (state & 0x1f) {
252 // Do not walk the whole buffer just to decode slice header
253 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
254 /* IDR or disposable slice
255 * No need to decode many bytes because MMCOs shall not be present. */
259 /* To decode up to MMCOs */
260 if (src_length > 1000)
265 ptr = ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
266 if (ptr == NULL || dst_length < 0)
269 init_get_bits(&h->gb, ptr, 8 * dst_length);
270 switch (h->nal_unit_type) {
272 ff_h264_decode_seq_parameter_set(h);
275 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
278 ff_h264_decode_sei(h);
283 h->prev_frame_num = 0;
284 h->prev_frame_num_offset = 0;
289 get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
290 slice_type = get_ue_golomb_31(&h->gb);
291 s->pict_type = golomb_to_pict_type[slice_type % 5];
292 if (h->sei_recovery_frame_cnt >= 0) {
293 /* key frame, since recovery_frame_cnt is set */
296 pps_id = get_ue_golomb(&h->gb);
297 if (pps_id >= MAX_PPS_COUNT) {
298 av_log(h->avctx, AV_LOG_ERROR,
299 "pps_id %u out of range\n", pps_id);
302 if (!h->pps_buffers[pps_id]) {
303 av_log(h->avctx, AV_LOG_ERROR,
304 "non-existing PPS %u referenced\n", pps_id);
307 h->pps = *h->pps_buffers[pps_id];
308 if (!h->sps_buffers[h->pps.sps_id]) {
309 av_log(h->avctx, AV_LOG_ERROR,
310 "non-existing SPS %u referenced\n", h->pps.sps_id);
313 h->sps = *h->sps_buffers[h->pps.sps_id];
314 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
316 if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
319 avctx->profile = ff_h264_get_profile(&h->sps);
320 avctx->level = h->sps.level_idc;
322 if (h->sps.frame_mbs_only_flag) {
323 h->picture_structure = PICT_FRAME;
325 if (get_bits1(&h->gb)) { // field_pic_flag
326 h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
328 h->picture_structure = PICT_FRAME;
332 if (h->nal_unit_type == NAL_IDR_SLICE)
333 get_ue_golomb(&h->gb); /* idr_pic_id */
334 if (h->sps.poc_type == 0) {
335 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
337 if (h->pps.pic_order_present == 1 &&
338 h->picture_structure == PICT_FRAME)
339 h->delta_poc_bottom = get_se_golomb(&h->gb);
342 if (h->sps.poc_type == 1 &&
343 !h->sps.delta_pic_order_always_zero_flag) {
344 h->delta_poc[0] = get_se_golomb(&h->gb);
346 if (h->pps.pic_order_present == 1 &&
347 h->picture_structure == PICT_FRAME)
348 h->delta_poc[1] = get_se_golomb(&h->gb);
351 /* Decode POC of this picture.
352 * The prev_ values needed for decoding POC of the next picture are not set here. */
353 field_poc[0] = field_poc[1] = INT_MAX;
354 ff_init_poc(h, field_poc, &s->output_picture_number);
356 /* Continue parsing to check if MMCO_RESET is present.
357 * FIXME: MMCO_RESET could appear in non-first slice.
358 * Maybe, we should parse all undisposable non-IDR slice of this
359 * picture until encountering MMCO_RESET in a slice of it. */
360 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
361 got_reset = scan_mmco_reset(s);
366 /* Set up the prev_ values for decoding POC of the next picture. */
367 h->prev_frame_num = got_reset ? 0 : h->frame_num;
368 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
369 if (h->nal_ref_idc != 0) {
371 h->prev_poc_msb = h->poc_msb;
372 h->prev_poc_lsb = h->poc_lsb;
376 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
380 if (h->sps.pic_struct_present_flag) {
381 switch (h->sei_pic_struct) {
382 case SEI_PIC_STRUCT_TOP_FIELD:
383 case SEI_PIC_STRUCT_BOTTOM_FIELD:
386 case SEI_PIC_STRUCT_FRAME:
387 case SEI_PIC_STRUCT_TOP_BOTTOM:
388 case SEI_PIC_STRUCT_BOTTOM_TOP:
391 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
392 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
395 case SEI_PIC_STRUCT_FRAME_DOUBLING:
398 case SEI_PIC_STRUCT_FRAME_TRIPLING:
402 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
406 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
409 if (h->picture_structure == PICT_FRAME) {
410 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
411 if (h->sps.pic_struct_present_flag) {
412 switch (h->sei_pic_struct) {
413 case SEI_PIC_STRUCT_TOP_BOTTOM:
414 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
415 s->field_order = AV_FIELD_TT;
417 case SEI_PIC_STRUCT_BOTTOM_TOP:
418 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
419 s->field_order = AV_FIELD_BB;
422 s->field_order = AV_FIELD_PROGRESSIVE;
426 if (field_poc[0] < field_poc[1])
427 s->field_order = AV_FIELD_TT;
428 else if (field_poc[0] > field_poc[1])
429 s->field_order = AV_FIELD_BB;
431 s->field_order = AV_FIELD_PROGRESSIVE;
434 if (h->picture_structure == PICT_TOP_FIELD)
435 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
437 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
438 s->field_order = AV_FIELD_UNKNOWN;
441 return 0; /* no need to evaluate the rest */
443 buf += h->is_avc ? nalsize : consumed;
447 /* didn't find a picture! */
448 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
452 static int h264_parse(AVCodecParserContext *s,
453 AVCodecContext *avctx,
454 const uint8_t **poutbuf, int *poutbuf_size,
455 const uint8_t *buf, int buf_size)
457 H264Context *h = s->priv_data;
458 ParseContext *pc = &h->parse_context;
463 if (avctx->extradata_size) {
465 // must be done like in decoder, otherwise opening the parser,
466 // letting it create extradata and then closing and opening again
467 // will cause has_b_frames to be always set.
468 // Note that estimate_timings_from_pts does exactly this.
469 if (!avctx->has_b_frames)
471 ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
475 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
478 next = h264_find_frame_end(h, buf, buf_size);
480 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
486 if (next < 0 && next != END_NOT_FOUND) {
487 av_assert1(pc->last_index + next >= 0);
488 h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
492 parse_nal_units(s, avctx, buf, buf_size);
494 if (h->sei_cpb_removal_delay >= 0) {
495 s->dts_sync_point = h->sei_buffering_period_present;
496 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
497 s->pts_dts_delta = h->sei_dpb_output_delay;
499 s->dts_sync_point = INT_MIN;
500 s->dts_ref_dts_delta = INT_MIN;
501 s->pts_dts_delta = INT_MIN;
504 if (s->flags & PARSER_FLAG_ONCE) {
505 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
509 *poutbuf_size = buf_size;
513 static int h264_split(AVCodecContext *avctx,
514 const uint8_t *buf, int buf_size)
520 for (i = 0; i <= buf_size; i++) {
521 if ((state & 0xFFFFFF1F) == 0x107)
523 /* if ((state&0xFFFFFF1F) == 0x101 ||
524 * (state&0xFFFFFF1F) == 0x102 ||
525 * (state&0xFFFFFF1F) == 0x105) {
528 if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x107 &&
529 (state & 0xFFFFFF1F) != 0x108 && (state & 0xFFFFFF1F) != 0x109) {
531 while (i > 4 && buf[i - 5] == 0)
537 state = (state << 8) | buf[i];
542 static void close(AVCodecParserContext *s)
544 H264Context *h = s->priv_data;
545 ParseContext *pc = &h->parse_context;
548 ff_h264_free_context(h);
551 static av_cold int init(AVCodecParserContext *s)
553 H264Context *h = s->priv_data;
554 h->thread_context[0] = h;
555 h->slice_context_count = 1;
556 ff_h264dsp_init(&h->h264dsp, 8, 1);
560 AVCodecParser ff_h264_parser = {
561 .codec_ids = { AV_CODEC_ID_H264 },
562 .priv_data_size = sizeof(H264Context),
564 .parser_parse = h264_parse,
565 .parser_close = close,