git.sesse.net Git - ffmpeg/blob - libavcodec/h264_parser.c

   1 /*
   2  * H.26L/H.264/AVC/JVT/14496-10/... parser
   3  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
   4  *
   5  * This file is part of Libav.
   6  *
   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.
  11  *
  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.
  16  *
  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
  20  */
  21
  22 /**
  23  * @file
  24  * H.264 / AVC / MPEG4 part10 parser.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #include "parser.h"
  29 #include "h264data.h"
  30 #include "golomb.h"
  31
  32 #include <assert.h>
  33
  34
  35 static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_size)
  36 {
  37     int i;
  38     uint32_t state;
  39     ParseContext *pc = &(h->s.parse_context);
  40 //printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]);
  41 //    mb_addr= pc->mb_addr - 1;
  42     state= pc->state;
  43     if(state>13)
  44         state= 7;
  45
  46     for(i=0; i<buf_size; i++){
  47         if(state==7){
  48 #if HAVE_FAST_UNALIGNED
  49         /* we check i<buf_size instead of i+3/7 because its simpler
  50          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  51          */
  52 #    if HAVE_FAST_64BIT
  53             while(i<buf_size && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  54                 i+=8;
  55 #    else
  56             while(i<buf_size && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  57                 i+=4;
  58 #    endif
  59 #endif
  60             for(; i<buf_size; i++){
  61                 if(!buf[i]){
  62                     state=2;
  63                     break;
  64                 }
  65             }
  66         }else if(state<=2){
  67             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  68             else if(buf[i]) state = 7;
  69             else            state>>=1; //2->1, 1->0, 0->0
  70         }else if(state<=5){
  71             int v= buf[i] & 0x1F;
  72             if(v==6 || v==7 || v==8 || v==9){
  73                 if(pc->frame_start_found){
  74                     i++;
  75                     goto found;
  76                 }
  77             }else if(v==1 || v==2 || v==5){
  78                 if(pc->frame_start_found){
  79                     state+=8;
  80                     continue;
  81                 }else
  82                     pc->frame_start_found = 1;
  83             }
  84             state= 7;
  85         }else{
  86             if(buf[i] & 0x80)
  87                 goto found;
  88             state= 7;
  89         }
  90     }
  91     pc->state= state;
  92     return END_NOT_FOUND;
  93
  94 found:
  95     pc->state=7;
  96     pc->frame_start_found= 0;
  97     return i-(state&5);
  98 }
  99
 100 /**
 101  * Parse NAL units of found picture and decode some basic information.
 102  *
 103  * @param s parser context.
 104  * @param avctx codec context.
 105  * @param buf buffer with field/frame data.
 106  * @param buf_size size of the buffer.
 107  */
 108 static inline int parse_nal_units(AVCodecParserContext *s,
 109                                   AVCodecContext *avctx,
 110                                   const uint8_t *buf, int buf_size)
 111 {
 112     H264Context *h = s->priv_data;
 113     const uint8_t *buf_end = buf + buf_size;
 114     unsigned int pps_id;
 115     unsigned int slice_type;
 116     int state = -1;
 117     const uint8_t *ptr;
 118
 119     /* set some sane default values */
 120     s->pict_type = AV_PICTURE_TYPE_I;
 121     s->key_frame = 0;
 122
 123     h->s.avctx= avctx;
 124     h->sei_recovery_frame_cnt = -1;
 125     h->sei_dpb_output_delay         =  0;
 126     h->sei_cpb_removal_delay        = -1;
 127     h->sei_buffering_period_present =  0;
 128
 129     if (!buf_size)
 130         return 0;
 131
 132     for(;;) {
 133         int src_length, dst_length, consumed;
 134         buf = avpriv_mpv_find_start_code(buf, buf_end, &state);
 135         if(buf >= buf_end)
 136             break;
 137         --buf;
 138         src_length = buf_end - buf;
 139         switch (state & 0x1f) {
 140         case NAL_SLICE:
 141         case NAL_IDR_SLICE:
 142             // Do not walk the whole buffer just to decode slice header
 143             if (src_length > 20)
 144                 src_length = 20;
 145             break;
 146         }
 147         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 148         if (ptr==NULL || dst_length < 0)
 149             break;
 150
 151         init_get_bits(&h->s.gb, ptr, 8*dst_length);
 152         switch(h->nal_unit_type) {
 153         case NAL_SPS:
 154             ff_h264_decode_seq_parameter_set(h);
 155             break;
 156         case NAL_PPS:
 157             ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
 158             break;
 159         case NAL_SEI:
 160             ff_h264_decode_sei(h);
 161             break;
 162         case NAL_IDR_SLICE:
 163             s->key_frame = 1;
 164             /* fall through */
 165         case NAL_SLICE:
 166             get_ue_golomb(&h->s.gb);  // skip first_mb_in_slice
 167             slice_type = get_ue_golomb_31(&h->s.gb);
 168             s->pict_type = golomb_to_pict_type[slice_type % 5];
 169             if (h->sei_recovery_frame_cnt >= 0) {
 170                 /* key frame, since recovery_frame_cnt is set */
 171                 s->key_frame = 1;
 172             }
 173             pps_id= get_ue_golomb(&h->s.gb);
 174             if(pps_id>=MAX_PPS_COUNT) {
 175                 av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
 176                 return -1;
 177             }
 178             if(!h->pps_buffers[pps_id]) {
 179                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 180                 return -1;
 181             }
 182             h->pps= *h->pps_buffers[pps_id];
 183             if(!h->sps_buffers[h->pps.sps_id]) {
 184                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 185                 return -1;
 186             }
 187             h->sps = *h->sps_buffers[h->pps.sps_id];
 188             h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
 189
 190             avctx->profile = ff_h264_get_profile(&h->sps);
 191             avctx->level   = h->sps.level_idc;
 192
 193             if(h->sps.frame_mbs_only_flag){
 194                 h->s.picture_structure= PICT_FRAME;
 195             }else{
 196                 if(get_bits1(&h->s.gb)) { //field_pic_flag
 197                     h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
 198                 } else {
 199                     h->s.picture_structure= PICT_FRAME;
 200                 }
 201             }
 202
 203             if(h->sps.pic_struct_present_flag) {
 204                 switch (h->sei_pic_struct) {
 205                     case SEI_PIC_STRUCT_TOP_FIELD:
 206                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 207                         s->repeat_pict = 0;
 208                         break;
 209                     case SEI_PIC_STRUCT_FRAME:
 210                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 211                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 212                         s->repeat_pict = 1;
 213                         break;
 214                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 215                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 216                         s->repeat_pict = 2;
 217                         break;
 218                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 219                         s->repeat_pict = 3;
 220                         break;
 221                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 222                         s->repeat_pict = 5;
 223                         break;
 224                     default:
 225                         s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 226                         break;
 227                 }
 228             } else {
 229                 s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 230             }
 231
 232             return 0; /* no need to evaluate the rest */
 233         }
 234         buf += consumed;
 235     }
 236     /* didn't find a picture! */
 237     av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit\n");
 238     return -1;
 239 }
 240
 241 static int h264_parse(AVCodecParserContext *s,
 242                       AVCodecContext *avctx,
 243                       const uint8_t **poutbuf, int *poutbuf_size,
 244                       const uint8_t *buf, int buf_size)
 245 {
 246     H264Context *h = s->priv_data;
 247     ParseContext *pc = &h->s.parse_context;
 248     int next;
 249
 250     if (!h->got_first) {
 251         h->got_first = 1;
 252         if (avctx->extradata_size) {
 253             h->s.avctx = avctx;
 254             // must be done like in the decoder.
 255             // otherwise opening the parser, creating extradata,
 256             // and then closing and opening again
 257             // will cause has_b_frames to be always set.
 258             // NB: estimate_timings_from_pts behaves exactly like this.
 259             if (!avctx->has_b_frames)
 260                 h->s.low_delay = 1;
 261             ff_h264_decode_extradata(h);
 262         }
 263     }
 264
 265     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 266         next= buf_size;
 267     }else{
 268         next= ff_h264_find_frame_end(h, buf, buf_size);
 269
 270         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 271             *poutbuf = NULL;
 272             *poutbuf_size = 0;
 273             return buf_size;
 274         }
 275
 276         if(next<0 && next != END_NOT_FOUND){
 277             assert(pc->last_index + next >= 0 );
 278             ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state
 279         }
 280     }
 281
 282     parse_nal_units(s, avctx, buf, buf_size);
 283
 284     if (h->sei_cpb_removal_delay >= 0) {
 285         s->dts_sync_point    = h->sei_buffering_period_present;
 286         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 287         s->pts_dts_delta     = h->sei_dpb_output_delay;
 288     } else {
 289         s->dts_sync_point    = INT_MIN;
 290         s->dts_ref_dts_delta = INT_MIN;
 291         s->pts_dts_delta     = INT_MIN;
 292     }
 293
 294     if (s->flags & PARSER_FLAG_ONCE) {
 295         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 296     }
 297
 298     *poutbuf = buf;
 299     *poutbuf_size = buf_size;
 300     return next;
 301 }
 302
 303 static int h264_split(AVCodecContext *avctx,
 304                       const uint8_t *buf, int buf_size)
 305 {
 306     int i;
 307     uint32_t state = -1;
 308     int has_sps= 0;
 309
 310     for(i=0; i<=buf_size; i++){
 311         if((state&0xFFFFFF1F) == 0x107)
 312             has_sps=1;
 313 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 314         }*/
 315         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 316             if(has_sps){
 317                 while(i>4 && buf[i-5]==0) i--;
 318                 return i-4;
 319             }
 320         }
 321         if (i<buf_size)
 322             state= (state<<8) | buf[i];
 323     }
 324     return 0;
 325 }
 326
 327 static void close(AVCodecParserContext *s)
 328 {
 329     H264Context *h = s->priv_data;
 330     ParseContext *pc = &h->s.parse_context;
 331
 332     av_free(pc->buffer);
 333     ff_h264_free_context(h);
 334 }
 335
 336 static int init(AVCodecParserContext *s)
 337 {
 338     H264Context *h = s->priv_data;
 339     h->thread_context[0] = h;
 340     h->s.slice_context_count = 1;
 341     return 0;
 342 }
 343
 344 AVCodecParser ff_h264_parser = {
 345     .codec_ids      = { CODEC_ID_H264 },
 346     .priv_data_size = sizeof(H264Context),
 347     .parser_init    = init,
 348     .parser_parse   = h264_parse,
 349     .parser_close   = close,
 350     .split          = h264_split,
 351 };