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, int buf_size)
41 ParseContext *pc = &h->parse_context;
42 // mb_addr= pc->mb_addr - 1;
47 for(i=0; i<buf_size; i++){
49 #if HAVE_FAST_UNALIGNED
50 /* we check i<buf_size instead of i+3/7 because its simpler
51 * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
54 while(i<buf_size && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
57 while(i<buf_size && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
61 for(; i<buf_size; i++){
68 if(buf[i]==1) state^= 5; //2->7, 1->4, 0->5
69 else if(buf[i]) state = 7;
70 else state>>=1; //2->1, 1->0, 0->0
73 if(v==6 || v==7 || v==8 || v==9){
74 if(pc->frame_start_found){
78 }else if(v==1 || v==2 || v==5){
79 if(pc->frame_start_found){
83 pc->frame_start_found = 1;
97 pc->frame_start_found= 0;
102 * Parse NAL units of found picture and decode some basic information.
104 * @param s parser context.
105 * @param avctx codec context.
106 * @param buf buffer with field/frame data.
107 * @param buf_size size of the buffer.
109 static inline int parse_nal_units(AVCodecParserContext *s,
110 AVCodecContext *avctx,
111 const uint8_t *buf, int buf_size)
113 H264Context *h = s->priv_data;
114 const uint8_t *buf_end = buf + buf_size;
116 unsigned int slice_type;
120 /* set some sane default values */
121 s->pict_type = AV_PICTURE_TYPE_I;
125 h->sei_recovery_frame_cnt = -1;
126 h->sei_dpb_output_delay = 0;
127 h->sei_cpb_removal_delay = -1;
128 h->sei_buffering_period_present = 0;
134 int src_length, dst_length, consumed;
135 buf = avpriv_find_start_code(buf, buf_end, &state);
139 src_length = buf_end - buf;
140 switch (state & 0x1f) {
143 // Do not walk the whole buffer just to decode slice header
148 ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
149 if (ptr==NULL || dst_length < 0)
152 init_get_bits(&h->gb, ptr, 8*dst_length);
153 switch(h->nal_unit_type) {
155 ff_h264_decode_seq_parameter_set(h);
158 ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
161 ff_h264_decode_sei(h);
167 get_ue_golomb(&h->gb); // skip first_mb_in_slice
168 slice_type = get_ue_golomb_31(&h->gb);
169 s->pict_type = golomb_to_pict_type[slice_type % 5];
170 if (h->sei_recovery_frame_cnt >= 0) {
171 /* key frame, since recovery_frame_cnt is set */
174 pps_id= get_ue_golomb(&h->gb);
175 if(pps_id>=MAX_PPS_COUNT) {
176 av_log(h->avctx, AV_LOG_ERROR, "pps_id out of range\n");
179 if(!h->pps_buffers[pps_id]) {
180 av_log(h->avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
183 h->pps= *h->pps_buffers[pps_id];
184 if(!h->sps_buffers[h->pps.sps_id]) {
185 av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
188 h->sps = *h->sps_buffers[h->pps.sps_id];
189 h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
191 avctx->profile = ff_h264_get_profile(&h->sps);
192 avctx->level = h->sps.level_idc;
194 if(h->sps.frame_mbs_only_flag){
195 h->picture_structure= PICT_FRAME;
197 if(get_bits1(&h->gb)) { //field_pic_flag
198 h->picture_structure= PICT_TOP_FIELD + get_bits1(&h->gb); //bottom_field_flag
200 h->picture_structure= PICT_FRAME;
204 if(h->sps.pic_struct_present_flag) {
205 switch (h->sei_pic_struct) {
206 case SEI_PIC_STRUCT_TOP_FIELD:
207 case SEI_PIC_STRUCT_BOTTOM_FIELD:
210 case SEI_PIC_STRUCT_FRAME:
211 case SEI_PIC_STRUCT_TOP_BOTTOM:
212 case SEI_PIC_STRUCT_BOTTOM_TOP:
215 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
216 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
219 case SEI_PIC_STRUCT_FRAME_DOUBLING:
222 case SEI_PIC_STRUCT_FRAME_TRIPLING:
226 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
230 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
233 return 0; /* no need to evaluate the rest */
237 /* didn't find a picture! */
238 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n");
242 static int h264_parse(AVCodecParserContext *s,
243 AVCodecContext *avctx,
244 const uint8_t **poutbuf, int *poutbuf_size,
245 const uint8_t *buf, int buf_size)
247 H264Context *h = s->priv_data;
248 ParseContext *pc = &h->parse_context;
253 if (avctx->extradata_size) {
255 // must be done like in the decoder.
256 // otherwise opening the parser, creating extradata,
257 // and then closing and opening again
258 // will cause has_b_frames to be always set.
259 // NB: estimate_timings_from_pts behaves exactly like this.
260 if (!avctx->has_b_frames)
262 ff_h264_decode_extradata(h);
266 if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
269 next = h264_find_frame_end(h, buf, buf_size);
271 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
277 if(next<0 && next != END_NOT_FOUND){
278 assert(pc->last_index + next >= 0 );
279 h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
283 parse_nal_units(s, avctx, buf, buf_size);
285 if (h->sei_cpb_removal_delay >= 0) {
286 s->dts_sync_point = h->sei_buffering_period_present;
287 s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
288 s->pts_dts_delta = h->sei_dpb_output_delay;
290 s->dts_sync_point = INT_MIN;
291 s->dts_ref_dts_delta = INT_MIN;
292 s->pts_dts_delta = INT_MIN;
295 if (s->flags & PARSER_FLAG_ONCE) {
296 s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
300 *poutbuf_size = buf_size;
304 static int h264_split(AVCodecContext *avctx,
305 const uint8_t *buf, int buf_size)
311 for(i=0; i<=buf_size; i++){
312 if((state&0xFFFFFF1F) == 0x107)
314 /* if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
316 if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
318 while(i>4 && buf[i-5]==0) i--;
323 state= (state<<8) | buf[i];
328 static void close(AVCodecParserContext *s)
330 H264Context *h = s->priv_data;
331 ParseContext *pc = &h->parse_context;
334 ff_h264_free_context(h);
337 static av_cold int init(AVCodecParserContext *s)
339 H264Context *h = s->priv_data;
340 h->thread_context[0] = h;
341 h->slice_context_count = 1;
345 AVCodecParser ff_h264_parser = {
346 .codec_ids = { AV_CODEC_ID_H264 },
347 .priv_data_size = sizeof(H264Context),
349 .parser_parse = h264_parse,
350 .parser_close = close,