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"
37 static int h264_find_frame_end(H264Context *h, const uint8_t *buf,
42 ParseContext *pc = &h->parse_context;
43 int next_avc= h->is_avc ? 0 : buf_size;
45 // mb_addr= pc->mb_addr - 1;
50 if (h->is_avc && !h->nal_length_size)
51 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
53 for (i = 0; i < buf_size; i++) {
57 for (j = 0; j < h->nal_length_size; j++)
58 nalsize = (nalsize << 8) | buf[i++];
59 if (nalsize <= 0 || nalsize > buf_size - i) {
60 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
63 next_avc = i + nalsize;
68 i += h->h264dsp.h264_find_start_code_candidate(buf + i, next_avc - i);
71 } else if (state <= 2) {
73 state ^= 5; // 2->7, 1->4, 0->5
77 state >>= 1; // 2->1, 1->0, 0->0
78 } else if (state <= 5) {
79 int v = buf[i] & 0x1F;
80 if (v == 6 || v == 7 || v == 8 || v == 9) {
81 if (pc->frame_start_found) {
85 } else if (v == 1 || v == 2 || v == 5) {
91 h->parse_history[h->parse_history_count++]= buf[i];
92 if (h->parse_history_count>3) {
93 unsigned int mb, last_mb= h->parse_last_mb;
96 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
97 h->parse_history_count=0;
98 mb= get_ue_golomb_long(&gb);
99 last_mb= h->parse_last_mb;
100 h->parse_last_mb= mb;
101 if (pc->frame_start_found) {
105 pc->frame_start_found = 1;
113 return END_NOT_FOUND;
117 pc->frame_start_found = 0;
120 return i - (state & 5) - 3 * (state > 7);
123 static int scan_mmco_reset(AVCodecParserContext *s)
125 H264Context *h = s->priv_data;
127 h->slice_type_nos = s->pict_type & 3;
129 if (h->pps.redundant_pic_cnt_present)
130 get_ue_golomb(&h->gb); // redundant_pic_count
132 if (ff_set_ref_count(h) < 0)
133 return AVERROR_INVALIDDATA;
135 if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
137 for (list = 0; list < h->list_count; list++) {
138 if (get_bits1(&h->gb)) {
140 for (index = 0; ; index++) {
141 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
143 if (reordering_of_pic_nums_idc < 3)
144 get_ue_golomb(&h->gb);
145 else if (reordering_of_pic_nums_idc > 3) {
146 av_log(h->avctx, AV_LOG_ERROR,
147 "illegal reordering_of_pic_nums_idc %d\n",
148 reordering_of_pic_nums_idc);
149 return AVERROR_INVALIDDATA;
153 if (index >= h->ref_count[list]) {
154 av_log(h->avctx, AV_LOG_ERROR,
155 "reference count %d overflow\n", index);
156 return AVERROR_INVALIDDATA;
163 if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) ||
164 (h->pps.weighted_bipred_idc == 1 && h->slice_type_nos == AV_PICTURE_TYPE_B))
165 ff_pred_weight_table(h);
167 if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
169 for (i = 0; i < MAX_MMCO_COUNT; i++) {
170 MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
171 if (opcode > (unsigned) MMCO_LONG) {
172 av_log(h->avctx, AV_LOG_ERROR,
173 "illegal memory management control operation %d\n",
175 return AVERROR_INVALIDDATA;
177 if (opcode == MMCO_END)
179 else if (opcode == MMCO_RESET)
182 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
183 get_ue_golomb(&h->gb);
184 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
185 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
186 get_ue_golomb_31(&h->gb);
194 * Parse NAL units of found picture and decode some basic information.
196 * @param s parser context.
197 * @param avctx codec context.
198 * @param buf buffer with field/frame data.
199 * @param buf_size size of the buffer.
201 static inline int parse_nal_units(AVCodecParserContext *s,
202 AVCodecContext *avctx,
203 const uint8_t *buf, int buf_size)
205 H264Context *h = s->priv_data;
206 const uint8_t *buf_end = buf + buf_size;
208 unsigned int slice_type;
209 int state = -1, got_reset = 0;
211 int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
214 /* set some sane default values */
215 s->pict_type = AV_PICTURE_TYPE_I;
217 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
220 ff_h264_reset_sei(h);
221 h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
227 int src_length, dst_length, consumed, nalsize = 0;
230 if (h->nal_length_size >= buf_end - buf) break;
232 for (i = 0; i < h->nal_length_size; i++)
233 nalsize = (nalsize << 8) | *buf++;
234 if (nalsize <= 0 || nalsize > buf_end - buf) {
235 av_log(h->avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
238 src_length = nalsize;
240 buf = avpriv_find_start_code(buf, buf_end, &state);
244 src_length = buf_end - buf;
246 switch (state & 0x1f) {
249 // Do not walk the whole buffer just to decode slice header
250 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
251 /* IDR or disposable slice
252 * No need to decode many bytes because MMCOs shall not be present. */
256 /* To decode up to MMCOs */
257 if (src_length > 1000)
262 ptr = ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
263 if (ptr == NULL || dst_length < 0)
266 init_get_bits(&h->gb, ptr, 8 * dst_length);
267 switch (h->nal_unit_type) {
269 ff_h264_decode_seq_parameter_set(h);
272 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
275 ff_h264_decode_sei(h);
280 h->prev_frame_num = 0;
281 h->prev_frame_num_offset = 0;
286 get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
287 slice_type = get_ue_golomb_31(&h->gb);
288 s->pict_type = golomb_to_pict_type[slice_type % 5];
289 if (h->sei_recovery_frame_cnt >= 0) {
290 /* key frame, since recovery_frame_cnt is set */
293 pps_id = get_ue_golomb(&h->gb);
294 if (pps_id >= MAX_PPS_COUNT) {
295 av_log(h->avctx, AV_LOG_ERROR,
296 "pps_id %u out of range\n", pps_id);
299 if (!h->pps_buffers[pps_id]) {
300 av_log(h->avctx, AV_LOG_ERROR,
301 "non-existing PPS %u referenced\n", pps_id);
304 h->pps = *h->pps_buffers[pps_id];
305 if (!h->sps_buffers[h->pps.sps_id]) {
306 av_log(h->avctx, AV_LOG_ERROR,
307 "non-existing SPS %u referenced\n", h->pps.sps_id);
310 h->sps = *h->sps_buffers[h->pps.sps_id];
311 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
313 if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
316 avctx->profile = ff_h264_get_profile(&h->sps);
317 avctx->level = h->sps.level_idc;
319 if (h->sps.frame_mbs_only_flag) {
320 h->picture_structure = PICT_FRAME;
322 if (get_bits1(&h->gb)) { // field_pic_flag
323 h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
325 h->picture_structure = PICT_FRAME;
329 if (h->nal_unit_type == NAL_IDR_SLICE)
330 get_ue_golomb(&h->gb); /* idr_pic_id */
331 if (h->sps.poc_type == 0) {
332 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
334 if (h->pps.pic_order_present == 1 &&
335 h->picture_structure == PICT_FRAME)
336 h->delta_poc_bottom = get_se_golomb(&h->gb);
339 if (h->sps.poc_type == 1 &&
340 !h->sps.delta_pic_order_always_zero_flag) {
341 h->delta_poc[0] = get_se_golomb(&h->gb);
343 if (h->pps.pic_order_present == 1 &&
344 h->picture_structure == PICT_FRAME)
345 h->delta_poc[1] = get_se_golomb(&h->gb);
348 /* Decode POC of this picture.
349 * The prev_ values needed for decoding POC of the next picture are not set here. */
350 field_poc[0] = field_poc[1] = INT_MAX;
351 ff_init_poc(h, field_poc, &s->output_picture_number);
353 /* Continue parsing to check if MMCO_RESET is present.
354 * FIXME: MMCO_RESET could appear in non-first slice.
355 * Maybe, we should parse all undisposable non-IDR slice of this
356 * picture until encountering MMCO_RESET in a slice of it. */
357 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
358 got_reset = scan_mmco_reset(s);
363 /* Set up the prev_ values for decoding POC of the next picture. */
364 h->prev_frame_num = got_reset ? 0 : h->frame_num;
365 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
366 if (h->nal_ref_idc != 0) {
368 h->prev_poc_msb = h->poc_msb;
369 h->prev_poc_lsb = h->poc_lsb;
373 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
377 if (h->sps.pic_struct_present_flag) {
378 switch (h->sei_pic_struct) {
379 case SEI_PIC_STRUCT_TOP_FIELD:
380 case SEI_PIC_STRUCT_BOTTOM_FIELD:
383 case SEI_PIC_STRUCT_FRAME:
384 case SEI_PIC_STRUCT_TOP_BOTTOM:
385 case SEI_PIC_STRUCT_BOTTOM_TOP:
388 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
389 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
392 case SEI_PIC_STRUCT_FRAME_DOUBLING:
395 case SEI_PIC_STRUCT_FRAME_TRIPLING:
399 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
403 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
406 if (h->picture_structure == PICT_FRAME) {
407 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
408 if (h->sps.pic_struct_present_flag) {
409 switch (h->sei_pic_struct) {
410 case SEI_PIC_STRUCT_TOP_BOTTOM:
411 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
412 s->field_order = AV_FIELD_TT;
414 case SEI_PIC_STRUCT_BOTTOM_TOP:
415 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
416 s->field_order = AV_FIELD_BB;
419 s->field_order = AV_FIELD_PROGRESSIVE;
423 if (field_poc[0] < field_poc[1])
424 s->field_order = AV_FIELD_TT;
425 else if (field_poc[0] > field_poc[1])
426 s->field_order = AV_FIELD_BB;
428 s->field_order = AV_FIELD_PROGRESSIVE;
431 if (h->picture_structure == PICT_TOP_FIELD)
432 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
434 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
435 s->field_order = AV_FIELD_UNKNOWN;
438 return 0; /* no need to evaluate the rest */
440 buf += h->is_avc ? nalsize : consumed;
444 /* didn't find a picture! */
445 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
449 static int h264_parse(AVCodecParserContext *s,
450 AVCodecContext *avctx,
451 const uint8_t **poutbuf, int *poutbuf_size,
452 const uint8_t *buf, int buf_size)
454 H264Context *h = s->priv_data;
455 ParseContext *pc = &h->parse_context;
460 if (avctx->extradata_size) {
462 // must be done like in decoder, otherwise opening the parser,
463 // letting it create extradata and then closing and opening again
464 // will cause has_b_frames to be always set.
465 // Note that estimate_timings_from_pts does exactly this.
466 if (!avctx->has_b_frames)
468 ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
472 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
475 next = h264_find_frame_end(h, buf, buf_size);
477 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
483 if (next < 0 && next != END_NOT_FOUND) {
484 av_assert1(pc->last_index + next >= 0);
485 h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
489 parse_nal_units(s, avctx, buf, buf_size);
491 if (h->sei_cpb_removal_delay >= 0) {
492 s->dts_sync_point = h->sei_buffering_period_present;
493 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
494 s->pts_dts_delta = h->sei_dpb_output_delay;
496 s->dts_sync_point = INT_MIN;
497 s->dts_ref_dts_delta = INT_MIN;
498 s->pts_dts_delta = INT_MIN;
501 if (s->flags & PARSER_FLAG_ONCE) {
502 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
506 *poutbuf_size = buf_size;
510 static int h264_split(AVCodecContext *avctx,
511 const uint8_t *buf, int buf_size)
517 for (i = 0; i <= buf_size; i++) {
518 if ((state & 0xFFFFFF1F) == 0x107)
520 /* if ((state&0xFFFFFF1F) == 0x101 ||
521 * (state&0xFFFFFF1F) == 0x102 ||
522 * (state&0xFFFFFF1F) == 0x105) {
525 if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x107 &&
526 (state & 0xFFFFFF1F) != 0x108 && (state & 0xFFFFFF1F) != 0x109) {
528 while (i > 4 && buf[i - 5] == 0)
534 state = (state << 8) | buf[i];
539 static void close(AVCodecParserContext *s)
541 H264Context *h = s->priv_data;
542 ParseContext *pc = &h->parse_context;
545 ff_h264_free_context(h);
548 static av_cold int init(AVCodecParserContext *s)
550 H264Context *h = s->priv_data;
551 h->thread_context[0] = h;
552 h->slice_context_count = 1;
553 ff_h264dsp_init(&h->h264dsp, 8, 1);
557 AVCodecParser ff_h264_parser = {
558 .codec_ids = { AV_CODEC_ID_H264 },
559 .priv_data_size = sizeof(H264Context),
561 .parser_parse = h264_parse,
562 .parser_close = close,