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 Libav.
7 * Libav 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 * Libav 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 Libav; 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 #include "libavutil/attributes.h"
37 static int h264_find_frame_end(H264Context *h, const uint8_t *buf,
42 ParseContext *pc = &h->parse_context;
43 // mb_addr= pc->mb_addr - 1;
48 for (i = 0; i < buf_size; i++) {
50 i += h->h264dsp.h264_find_start_code_candidate(buf + i, buf_size - i);
53 } else if (state <= 2) {
55 state ^= 5; // 2->7, 1->4, 0->5
59 state >>= 1; // 2->1, 1->0, 0->0
60 } else if (state <= 5) {
61 int v = buf[i] & 0x1F;
62 if (v == 6 || v == 7 || v == 8 || v == 9) {
63 if (pc->frame_start_found) {
67 } else if (v == 1 || v == 2 || v == 5) {
68 if (pc->frame_start_found) {
72 pc->frame_start_found = 1;
86 pc->frame_start_found = 0;
87 return i - (state & 5);
90 static int scan_mmco_reset(AVCodecParserContext *s)
92 H264Context *h = s->priv_data;
94 h->slice_type_nos = s->pict_type & 3;
96 if (h->pps.redundant_pic_cnt_present)
97 get_ue_golomb(&h->gb); // redundant_pic_count
99 if (ff_set_ref_count(h) < 0)
100 return AVERROR_INVALIDDATA;
102 if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
104 for (list = 0; list < h->list_count; list++) {
105 if (get_bits1(&h->gb)) {
107 for (index = 0; ; index++) {
108 unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
110 if (reordering_of_pic_nums_idc < 3)
111 get_ue_golomb(&h->gb);
112 else if (reordering_of_pic_nums_idc > 3) {
113 av_log(h->avctx, AV_LOG_ERROR,
114 "illegal reordering_of_pic_nums_idc %d\n",
115 reordering_of_pic_nums_idc);
116 return AVERROR_INVALIDDATA;
120 if (index >= h->ref_count[list]) {
121 av_log(h->avctx, AV_LOG_ERROR,
122 "reference count %d overflow\n", index);
123 return AVERROR_INVALIDDATA;
130 if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) ||
131 (h->pps.weighted_bipred_idc == 1 && h->slice_type_nos == AV_PICTURE_TYPE_B))
132 ff_pred_weight_table(h);
134 if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
136 for (i = 0; i < MAX_MMCO_COUNT; i++) {
137 MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
138 if (opcode > (unsigned) MMCO_LONG) {
139 av_log(h->avctx, AV_LOG_ERROR,
140 "illegal memory management control operation %d\n",
142 return AVERROR_INVALIDDATA;
144 if (opcode == MMCO_END)
146 else if (opcode == MMCO_RESET)
149 if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
150 get_ue_golomb(&h->gb);
151 if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
152 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
153 get_ue_golomb_31(&h->gb);
161 * Parse NAL units of found picture and decode some basic information.
163 * @param s parser context.
164 * @param avctx codec context.
165 * @param buf buffer with field/frame data.
166 * @param buf_size size of the buffer.
168 static inline int parse_nal_units(AVCodecParserContext *s,
169 AVCodecContext *avctx,
170 const uint8_t *buf, int buf_size)
172 H264Context *h = s->priv_data;
173 const uint8_t *buf_end = buf + buf_size;
175 unsigned int slice_type;
176 int state = -1, got_reset = 0;
180 /* set some sane default values */
181 s->pict_type = AV_PICTURE_TYPE_I;
183 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
186 ff_h264_reset_sei(h);
192 int src_length, dst_length, consumed;
193 buf = avpriv_find_start_code(buf, buf_end, &state);
197 src_length = buf_end - buf;
198 switch (state & 0x1f) {
201 // Do not walk the whole buffer just to decode slice header
202 if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
203 /* IDR or disposable slice
204 * No need to decode many bytes because MMCOs shall not be present. */
208 /* To decode up to MMCOs */
209 if (src_length > 1000)
214 ptr = ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
215 if (ptr == NULL || dst_length < 0)
218 init_get_bits(&h->gb, ptr, 8 * dst_length);
219 switch (h->nal_unit_type) {
221 ff_h264_decode_seq_parameter_set(h);
224 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
227 ff_h264_decode_sei(h);
232 h->prev_frame_num = 0;
233 h->prev_frame_num_offset = 0;
238 get_ue_golomb(&h->gb); // skip first_mb_in_slice
239 slice_type = get_ue_golomb_31(&h->gb);
240 s->pict_type = golomb_to_pict_type[slice_type % 5];
241 if (h->sei_recovery_frame_cnt >= 0) {
242 /* key frame, since recovery_frame_cnt is set */
245 pps_id = get_ue_golomb(&h->gb);
246 if (pps_id >= MAX_PPS_COUNT) {
247 av_log(h->avctx, AV_LOG_ERROR,
248 "pps_id %u out of range\n", pps_id);
251 if (!h->pps_buffers[pps_id]) {
252 av_log(h->avctx, AV_LOG_ERROR,
253 "non-existing PPS %u referenced\n", pps_id);
256 h->pps = *h->pps_buffers[pps_id];
257 if (!h->sps_buffers[h->pps.sps_id]) {
258 av_log(h->avctx, AV_LOG_ERROR,
259 "non-existing SPS %u referenced\n", h->pps.sps_id);
262 h->sps = *h->sps_buffers[h->pps.sps_id];
263 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
265 avctx->profile = ff_h264_get_profile(&h->sps);
266 avctx->level = h->sps.level_idc;
268 if (h->sps.frame_mbs_only_flag) {
269 h->picture_structure = PICT_FRAME;
271 if (get_bits1(&h->gb)) { // field_pic_flag
272 h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
274 h->picture_structure = PICT_FRAME;
278 if (h->nal_unit_type == NAL_IDR_SLICE)
279 get_ue_golomb(&h->gb); /* idr_pic_id */
280 if (h->sps.poc_type == 0) {
281 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
283 if (h->pps.pic_order_present == 1 &&
284 h->picture_structure == PICT_FRAME)
285 h->delta_poc_bottom = get_se_golomb(&h->gb);
288 if (h->sps.poc_type == 1 &&
289 !h->sps.delta_pic_order_always_zero_flag) {
290 h->delta_poc[0] = get_se_golomb(&h->gb);
292 if (h->pps.pic_order_present == 1 &&
293 h->picture_structure == PICT_FRAME)
294 h->delta_poc[1] = get_se_golomb(&h->gb);
297 /* Decode POC of this picture.
298 * The prev_ values needed for decoding POC of the next picture are not set here. */
299 field_poc[0] = field_poc[1] = INT_MAX;
300 ff_init_poc(h, field_poc, &s->output_picture_number);
302 /* Continue parsing to check if MMCO_RESET is present.
303 * FIXME: MMCO_RESET could appear in non-first slice.
304 * Maybe, we should parse all undisposable non-IDR slice of this
305 * picture until encountering MMCO_RESET in a slice of it. */
306 if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
307 got_reset = scan_mmco_reset(s);
312 /* Set up the prev_ values for decoding POC of the next picture. */
313 h->prev_frame_num = got_reset ? 0 : h->frame_num;
314 h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
315 if (h->nal_ref_idc != 0) {
317 h->prev_poc_msb = h->poc_msb;
318 h->prev_poc_lsb = h->poc_lsb;
322 h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
326 if (h->sps.pic_struct_present_flag) {
327 switch (h->sei_pic_struct) {
328 case SEI_PIC_STRUCT_TOP_FIELD:
329 case SEI_PIC_STRUCT_BOTTOM_FIELD:
332 case SEI_PIC_STRUCT_FRAME:
333 case SEI_PIC_STRUCT_TOP_BOTTOM:
334 case SEI_PIC_STRUCT_BOTTOM_TOP:
337 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
338 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
341 case SEI_PIC_STRUCT_FRAME_DOUBLING:
344 case SEI_PIC_STRUCT_FRAME_TRIPLING:
348 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
352 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
355 if (h->picture_structure == PICT_FRAME) {
356 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
357 if (h->sps.pic_struct_present_flag) {
358 switch (h->sei_pic_struct) {
359 case SEI_PIC_STRUCT_TOP_BOTTOM:
360 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
361 s->field_order = AV_FIELD_TT;
363 case SEI_PIC_STRUCT_BOTTOM_TOP:
364 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
365 s->field_order = AV_FIELD_BB;
368 s->field_order = AV_FIELD_PROGRESSIVE;
372 if (field_poc[0] < field_poc[1])
373 s->field_order = AV_FIELD_TT;
374 else if (field_poc[0] > field_poc[1])
375 s->field_order = AV_FIELD_BB;
377 s->field_order = AV_FIELD_PROGRESSIVE;
380 if (h->picture_structure == PICT_TOP_FIELD)
381 s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
383 s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
384 s->field_order = AV_FIELD_UNKNOWN;
387 return 0; /* no need to evaluate the rest */
391 /* didn't find a picture! */
392 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n");
396 static int h264_parse(AVCodecParserContext *s,
397 AVCodecContext *avctx,
398 const uint8_t **poutbuf, int *poutbuf_size,
399 const uint8_t *buf, int buf_size)
401 H264Context *h = s->priv_data;
402 ParseContext *pc = &h->parse_context;
407 if (avctx->extradata_size) {
409 // must be done like in the decoder.
410 // otherwise opening the parser, creating extradata,
411 // and then closing and opening again
412 // will cause has_b_frames to be always set.
413 // NB: estimate_timings_from_pts behaves exactly like this.
414 if (!avctx->has_b_frames)
416 ff_h264_decode_extradata(h);
420 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
423 next = h264_find_frame_end(h, buf, buf_size);
425 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
431 if (next < 0 && next != END_NOT_FOUND) {
432 assert(pc->last_index + next >= 0);
433 h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
437 parse_nal_units(s, avctx, buf, buf_size);
439 if (h->sei_cpb_removal_delay >= 0) {
440 s->dts_sync_point = h->sei_buffering_period_present;
441 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
442 s->pts_dts_delta = h->sei_dpb_output_delay;
444 s->dts_sync_point = INT_MIN;
445 s->dts_ref_dts_delta = INT_MIN;
446 s->pts_dts_delta = INT_MIN;
449 if (s->flags & PARSER_FLAG_ONCE) {
450 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
454 *poutbuf_size = buf_size;
458 static int h264_split(AVCodecContext *avctx,
459 const uint8_t *buf, int buf_size)
465 for (i = 0; i <= buf_size; i++) {
466 if ((state & 0xFFFFFF1F) == 0x107)
468 /* if((state&0xFFFFFF1F) == 0x101 ||
469 * (state&0xFFFFFF1F) == 0x102 ||
470 * (state&0xFFFFFF1F) == 0x105) {
473 if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x107 &&
474 (state & 0xFFFFFF1F) != 0x108 && (state & 0xFFFFFF1F) != 0x109) {
476 while (i > 4 && buf[i - 5] == 0)
482 state = (state << 8) | buf[i];
487 static void close(AVCodecParserContext *s)
489 H264Context *h = s->priv_data;
490 ParseContext *pc = &h->parse_context;
493 ff_h264_free_context(h);
496 static av_cold int init(AVCodecParserContext *s)
498 H264Context *h = s->priv_data;
499 h->thread_context[0] = h;
500 h->slice_context_count = 1;
501 ff_h264dsp_init(&h->h264dsp, 8, 1);
505 AVCodecParser ff_h264_parser = {
506 .codec_ids = { AV_CODEC_ID_H264 },
507 .priv_data_size = sizeof(H264Context),
509 .parser_parse = h264_parse,
510 .parser_close = close,