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.startcode_find_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>5) {
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 h->parse_last_mb= mb;
103 if (pc->frame_start_found) {
107 pc->frame_start_found = 1;
115 return END_NOT_FOUND;
119 pc->frame_start_found = 0;
122 return i - (state & 5) - 5 * (state > 7);
125 static int scan_mmco_reset(AVCodecParserContext *s)
127 H264Context *h = s->priv_data;
129 h->slice_type_nos = s->pict_type & 3;
131 if (h->pps.redundant_pic_cnt_present)
132 get_ue_golomb(&h->gb); // redundant_pic_count
134 if (ff_set_ref_count(h) < 0)
135 return AVERROR_INVALIDDATA;
137 if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
139 for (list = 0; list < h->list_count; list++) {
140 if (get_bits1(&h->gb)) {
142 for (index = 0; ; index++) {
143 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
145 if (reordering_of_pic_nums_idc < 3)
146 get_ue_golomb(&h->gb);
147 else if (reordering_of_pic_nums_idc > 3) {
148 av_log(h->avctx, AV_LOG_ERROR,
149 "illegal reordering_of_pic_nums_idc %d\n",
150 reordering_of_pic_nums_idc);
151 return AVERROR_INVALIDDATA;
155 if (index >= h->ref_count[list]) {
156 av_log(h->avctx, AV_LOG_ERROR,
157 "reference count %d overflow\n", index);
158 return AVERROR_INVALIDDATA;
165 if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) ||
166 (h->pps.weighted_bipred_idc == 1 && h->slice_type_nos == AV_PICTURE_TYPE_B))
167 ff_pred_weight_table(h);
169 if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
171 for (i = 0; i < MAX_MMCO_COUNT; i++) {
172 MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
173 if (opcode > (unsigned) MMCO_LONG) {
174 av_log(h->avctx, AV_LOG_ERROR,
175 "illegal memory management control operation %d\n",
177 return AVERROR_INVALIDDATA;
179 if (opcode == MMCO_END)
181 else if (opcode == MMCO_RESET)
184 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
185 get_ue_golomb(&h->gb);
186 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
187 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
188 get_ue_golomb_31(&h->gb);
196 * Parse NAL units of found picture and decode some basic information.
198 * @param s parser context.
199 * @param avctx codec context.
200 * @param buf buffer with field/frame data.
201 * @param buf_size size of the buffer.
203 static inline int parse_nal_units(AVCodecParserContext *s,
204 AVCodecContext *avctx,
205 const uint8_t * const buf, int buf_size)
207 H264Context *h = s->priv_data;
208 int buf_index, next_avc;
210 unsigned int slice_type;
211 int state = -1, got_reset = 0;
213 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
216 /* set some sane default values */
217 s->pict_type = AV_PICTURE_TYPE_I;
219 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
222 ff_h264_reset_sei(h);
223 h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
229 next_avc = h->is_avc ? 0 : buf_size;
231 int src_length, dst_length, consumed, nalsize = 0;
233 if (buf_index >= next_avc) {
234 nalsize = get_avc_nalsize(h, buf, buf_size, &buf_index);
237 next_avc = buf_index + nalsize;
239 buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
240 if (buf_index >= buf_size)
242 if (buf_index >= next_avc)
245 src_length = next_avc - buf_index;
247 state = buf[buf_index];
248 switch (state & 0x1f) {
251 // Do not walk the whole buffer just to decode slice header
252 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
253 /* IDR or disposable slice
254 * No need to decode many bytes because MMCOs shall not be present. */
258 /* To decode up to MMCOs */
259 if (src_length > 1000)
264 ptr = ff_h264_decode_nal(h, buf + buf_index, &dst_length,
265 &consumed, src_length);
266 if (!ptr || dst_length < 0)
269 buf_index += consumed;
271 init_get_bits(&h->gb, ptr, 8 * dst_length);
272 switch (h->nal_unit_type) {
274 ff_h264_decode_seq_parameter_set(h);
277 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
280 ff_h264_decode_sei(h);
285 h->prev_frame_num = 0;
286 h->prev_frame_num_offset = 0;
291 get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
292 slice_type = get_ue_golomb_31(&h->gb);
293 s->pict_type = golomb_to_pict_type[slice_type % 5];
294 if (h->sei_recovery_frame_cnt >= 0) {
295 /* key frame, since recovery_frame_cnt is set */
298 pps_id = get_ue_golomb(&h->gb);
299 if (pps_id >= MAX_PPS_COUNT) {
300 av_log(h->avctx, AV_LOG_ERROR,
301 "pps_id %u out of range\n", pps_id);
304 if (!h->pps_buffers[pps_id]) {
305 av_log(h->avctx, AV_LOG_ERROR,
306 "non-existing PPS %u referenced\n", pps_id);
309 h->pps = *h->pps_buffers[pps_id];
310 if (!h->sps_buffers[h->pps.sps_id]) {
311 av_log(h->avctx, AV_LOG_ERROR,
312 "non-existing SPS %u referenced\n", h->pps.sps_id);
315 h->sps = *h->sps_buffers[h->pps.sps_id];
316 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
318 if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
321 avctx->profile = ff_h264_get_profile(&h->sps);
322 avctx->level = h->sps.level_idc;
324 if (h->sps.frame_mbs_only_flag) {
325 h->picture_structure = PICT_FRAME;
327 if (get_bits1(&h->gb)) { // field_pic_flag
328 h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
330 h->picture_structure = PICT_FRAME;
334 if (h->nal_unit_type == NAL_IDR_SLICE)
335 get_ue_golomb(&h->gb); /* idr_pic_id */
336 if (h->sps.poc_type == 0) {
337 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
339 if (h->pps.pic_order_present == 1 &&
340 h->picture_structure == PICT_FRAME)
341 h->delta_poc_bottom = get_se_golomb(&h->gb);
344 if (h->sps.poc_type == 1 &&
345 !h->sps.delta_pic_order_always_zero_flag) {
346 h->delta_poc[0] = get_se_golomb(&h->gb);
348 if (h->pps.pic_order_present == 1 &&
349 h->picture_structure == PICT_FRAME)
350 h->delta_poc[1] = get_se_golomb(&h->gb);
353 /* Decode POC of this picture.
354 * The prev_ values needed for decoding POC of the next picture are not set here. */
355 field_poc[0] = field_poc[1] = INT_MAX;
356 ff_init_poc(h, field_poc, &s->output_picture_number);
358 /* Continue parsing to check if MMCO_RESET is present.
359 * FIXME: MMCO_RESET could appear in non-first slice.
360 * Maybe, we should parse all undisposable non-IDR slice of this
361 * picture until encountering MMCO_RESET in a slice of it. */
362 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
363 got_reset = scan_mmco_reset(s);
368 /* Set up the prev_ values for decoding POC of the next picture. */
369 h->prev_frame_num = got_reset ? 0 : h->frame_num;
370 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
371 if (h->nal_ref_idc != 0) {
373 h->prev_poc_msb = h->poc_msb;
374 h->prev_poc_lsb = h->poc_lsb;
378 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
382 if (h->sps.pic_struct_present_flag) {
383 switch (h->sei_pic_struct) {
384 case SEI_PIC_STRUCT_TOP_FIELD:
385 case SEI_PIC_STRUCT_BOTTOM_FIELD:
388 case SEI_PIC_STRUCT_FRAME:
389 case SEI_PIC_STRUCT_TOP_BOTTOM:
390 case SEI_PIC_STRUCT_BOTTOM_TOP:
393 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
394 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
397 case SEI_PIC_STRUCT_FRAME_DOUBLING:
400 case SEI_PIC_STRUCT_FRAME_TRIPLING:
404 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
408 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
411 if (h->picture_structure == PICT_FRAME) {
412 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
413 if (h->sps.pic_struct_present_flag) {
414 switch (h->sei_pic_struct) {
415 case SEI_PIC_STRUCT_TOP_BOTTOM:
416 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
417 s->field_order = AV_FIELD_TT;
419 case SEI_PIC_STRUCT_BOTTOM_TOP:
420 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
421 s->field_order = AV_FIELD_BB;
424 s->field_order = AV_FIELD_PROGRESSIVE;
428 if (field_poc[0] < field_poc[1])
429 s->field_order = AV_FIELD_TT;
430 else if (field_poc[0] > field_poc[1])
431 s->field_order = AV_FIELD_BB;
433 s->field_order = AV_FIELD_PROGRESSIVE;
436 if (h->picture_structure == PICT_TOP_FIELD)
437 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
439 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
440 s->field_order = AV_FIELD_UNKNOWN;
443 return 0; /* no need to evaluate the rest */
448 /* didn't find a picture! */
449 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
453 static int h264_parse(AVCodecParserContext *s,
454 AVCodecContext *avctx,
455 const uint8_t **poutbuf, int *poutbuf_size,
456 const uint8_t *buf, int buf_size)
458 H264Context *h = s->priv_data;
459 ParseContext *pc = &h->parse_context;
464 if (avctx->extradata_size) {
466 // must be done like in decoder, otherwise opening the parser,
467 // letting it create extradata and then closing and opening again
468 // will cause has_b_frames to be always set.
469 // Note that estimate_timings_from_pts does exactly this.
470 if (!avctx->has_b_frames)
472 ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
476 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
479 next = h264_find_frame_end(h, buf, buf_size);
481 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
487 if (next < 0 && next != END_NOT_FOUND) {
488 av_assert1(pc->last_index + next >= 0);
489 h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
493 parse_nal_units(s, avctx, buf, buf_size);
495 if (avctx->framerate.num)
496 avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
497 if (h->sei_cpb_removal_delay >= 0) {
498 s->dts_sync_point = h->sei_buffering_period_present;
499 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
500 s->pts_dts_delta = h->sei_dpb_output_delay;
502 s->dts_sync_point = INT_MIN;
503 s->dts_ref_dts_delta = INT_MIN;
504 s->pts_dts_delta = INT_MIN;
507 if (s->flags & PARSER_FLAG_ONCE) {
508 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
512 *poutbuf_size = buf_size;
516 static int h264_split(AVCodecContext *avctx,
517 const uint8_t *buf, int buf_size)
524 for (i = 0; i <= buf_size; i++) {
525 if ((state & 0xFFFFFF1F) == 0x107)
527 if ((state & 0xFFFFFF1F) == 0x108)
529 /* if ((state&0xFFFFFF1F) == 0x101 ||
530 * (state&0xFFFFFF1F) == 0x102 ||
531 * (state&0xFFFFFF1F) == 0x105) {
534 if ((state & 0xFFFFFF00) == 0x100 && ((state & 0xFFFFFF1F) != 0x106 || has_pps) &&
535 (state & 0xFFFFFF1F) != 0x107 && (state & 0xFFFFFF1F) != 0x108 &&
536 (state & 0xFFFFFF1F) != 0x109 && (state & 0xFFFFFF1F) != 0x10d &&
537 (state & 0xFFFFFF1F) != 0x10f) {
539 while (i > 4 && buf[i - 5] == 0)
545 state = (state << 8) | buf[i];
550 static void close(AVCodecParserContext *s)
552 H264Context *h = s->priv_data;
553 ParseContext *pc = &h->parse_context;
555 av_freep(&pc->buffer);
556 ff_h264_free_context(h);
559 static av_cold int init(AVCodecParserContext *s)
561 H264Context *h = s->priv_data;
562 h->thread_context[0] = h;
563 h->slice_context_count = 1;
564 ff_h264dsp_init(&h->h264dsp, 8, 1);
568 AVCodecParser ff_h264_parser = {
569 .codec_ids = { AV_CODEC_ID_H264 },
570 .priv_data_size = sizeof(H264Context),
572 .parser_parse = h264_parse,
573 .parser_close = close,