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 FFmpeg.
   6  *
   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.
  11  *
  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.
  16  *
  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
  20  */
  21
  22 /**
  23  * @file
  24  * H.264 / AVC / MPEG4 part10 parser.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #define UNCHECKED_BITSTREAM_READER 1
  29
  30 #include "parser.h"
  31 #include "h264data.h"
  32 #include "golomb.h"
  33
  34
  35 static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_size)
  36 {
  37     int i, j;
  38     uint32_t state;
  39     ParseContext *pc = &(h->s.parse_context);
  40     int next_avc= h->is_avc ? 0 : buf_size;
  41
  42 //    mb_addr= pc->mb_addr - 1;
  43     state= pc->state;
  44     if(state>13)
  45         state= 7;
  46
  47     if(h->is_avc && !h->nal_length_size)
  48         av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
  49
  50     for(i=0; i<buf_size; i++){
  51         if(i >= next_avc) {
  52             int nalsize = 0;
  53             i = next_avc;
  54             for(j = 0; j < h->nal_length_size; j++)
  55                 nalsize = (nalsize << 8) | buf[i++];
  56             if(nalsize <= 0 || nalsize > buf_size - i){
  57                 av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
  58                 return buf_size;
  59             }
  60             next_avc= i + nalsize;
  61             state= 5;
  62         }
  63
  64         if(state==7){
  65 #if HAVE_FAST_UNALIGNED
  66         /* we check i<buf_size instead of i+3/7 because its simpler
  67          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  68          */
  69 #    if HAVE_FAST_64BIT
  70             while(i<next_avc && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  71                 i+=8;
  72 #    else
  73             while(i<next_avc && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  74                 i+=4;
  75 #    endif
  76 #endif
  77             for(; i<next_avc; i++){
  78                 if(!buf[i]){
  79                     state=2;
  80                     break;
  81                 }
  82             }
  83         }else if(state<=2){
  84             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  85             else if(buf[i]) state = 7;
  86             else            state>>=1; //2->1, 1->0, 0->0
  87         }else if(state<=5){
  88             int v= buf[i] & 0x1F;
  89             if(v==6 || v==7 || v==8 || v==9){
  90                 if(pc->frame_start_found){
  91                     i++;
  92                     goto found;
  93                 }
  94             }else if(v==1 || v==2 || v==5){
  95                 state+=8;
  96                 continue;
  97             }
  98             state= 7;
  99         }else{
 100             h->parse_history[h->parse_history_count++]= buf[i];
 101             if(h->parse_history_count>3){
 102                 unsigned int mb, last_mb= h->parse_last_mb;
 103                 GetBitContext gb;
 104
 105                 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
 106                 h->parse_history_count=0;
 107                 mb= get_ue_golomb_long(&gb);
 108                 last_mb= h->parse_last_mb;
 109                 h->parse_last_mb= mb;
 110                 if(pc->frame_start_found){
 111                     if(mb <= last_mb)
 112                         goto found;
 113                 }else
 114                     pc->frame_start_found = 1;
 115                 state= 7;
 116             }
 117         }
 118     }
 119     pc->state= state;
 120     if(h->is_avc)
 121         return next_avc;
 122     return END_NOT_FOUND;
 123
 124 found:
 125     pc->state=7;
 126     pc->frame_start_found= 0;
 127     if(h->is_avc)
 128         return next_avc;
 129     return i-(state&5) - 3*(state>7);
 130 }
 131
 132 /**
 133  * Parse NAL units of found picture and decode some basic information.
 134  *
 135  * @param s parser context.
 136  * @param avctx codec context.
 137  * @param buf buffer with field/frame data.
 138  * @param buf_size size of the buffer.
 139  */
 140 static inline int parse_nal_units(AVCodecParserContext *s,
 141                                   AVCodecContext *avctx,
 142                                   const uint8_t *buf, int buf_size)
 143 {
 144     H264Context *h = s->priv_data;
 145     const uint8_t *buf_end = buf + buf_size;
 146     unsigned int pps_id;
 147     unsigned int slice_type;
 148     int state = -1;
 149     const uint8_t *ptr;
 150     int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
 151
 152     /* set some sane default values */
 153     s->pict_type = AV_PICTURE_TYPE_I;
 154     s->key_frame = 0;
 155
 156     h->s.avctx= avctx;
 157     h->sei_recovery_frame_cnt = -1;
 158     h->sei_dpb_output_delay         =  0;
 159     h->sei_cpb_removal_delay        = -1;
 160     h->sei_buffering_period_present =  0;
 161
 162     if (!buf_size)
 163         return 0;
 164
 165     for(;;) {
 166         int src_length, dst_length, consumed, nalsize = 0;
 167         if (h->is_avc) {
 168             int i;
 169             if (h->nal_length_size >= buf_end - buf) break;
 170             nalsize = 0;
 171             for (i = 0; i < h->nal_length_size; i++)
 172                 nalsize = (nalsize << 8) | *buf++;
 173             if (nalsize <= 0 || nalsize > buf_end - buf) {
 174                 av_log(h->s.avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
 175                 break;
 176             }
 177             src_length = nalsize;
 178         } else {
 179         buf = avpriv_mpv_find_start_code(buf, buf_end, &state);
 180         if(buf >= buf_end)
 181             break;
 182         --buf;
 183         src_length = buf_end - buf;
 184         }
 185         switch (state & 0x1f) {
 186         case NAL_SLICE:
 187         case NAL_IDR_SLICE:
 188             // Do not walk the whole buffer just to decode slice header
 189             if (src_length > 20)
 190                 src_length = 20;
 191             break;
 192         }
 193         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 194         if (ptr==NULL || dst_length < 0)
 195             break;
 196
 197         init_get_bits(&h->s.gb, ptr, 8*dst_length);
 198         switch(h->nal_unit_type) {
 199         case NAL_SPS:
 200             ff_h264_decode_seq_parameter_set(h);
 201             break;
 202         case NAL_PPS:
 203             ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
 204             break;
 205         case NAL_SEI:
 206             ff_h264_decode_sei(h);
 207             break;
 208         case NAL_IDR_SLICE:
 209             s->key_frame = 1;
 210             /* fall through */
 211         case NAL_SLICE:
 212             get_ue_golomb_long(&h->s.gb);  // skip first_mb_in_slice
 213             slice_type = get_ue_golomb_31(&h->s.gb);
 214             s->pict_type = golomb_to_pict_type[slice_type % 5];
 215             if (h->sei_recovery_frame_cnt >= 0) {
 216                 /* key frame, since recovery_frame_cnt is set */
 217                 s->key_frame = 1;
 218             }
 219             pps_id= get_ue_golomb(&h->s.gb);
 220             if(pps_id>=MAX_PPS_COUNT) {
 221                 av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
 222                 return -1;
 223             }
 224             if(!h->pps_buffers[pps_id]) {
 225                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 226                 return -1;
 227             }
 228             h->pps= *h->pps_buffers[pps_id];
 229             if(!h->sps_buffers[h->pps.sps_id]) {
 230                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 231                 return -1;
 232             }
 233             h->sps = *h->sps_buffers[h->pps.sps_id];
 234             h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
 235
 236             avctx->profile = ff_h264_get_profile(&h->sps);
 237             avctx->level   = h->sps.level_idc;
 238
 239             if(h->sps.frame_mbs_only_flag){
 240                 h->s.picture_structure= PICT_FRAME;
 241             }else{
 242                 if(get_bits1(&h->s.gb)) { //field_pic_flag
 243                     h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
 244                 } else {
 245                     h->s.picture_structure= PICT_FRAME;
 246                 }
 247             }
 248
 249             if(h->sps.pic_struct_present_flag) {
 250                 switch (h->sei_pic_struct) {
 251                     case SEI_PIC_STRUCT_TOP_FIELD:
 252                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 253                         s->repeat_pict = 0;
 254                         break;
 255                     case SEI_PIC_STRUCT_FRAME:
 256                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 257                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 258                         s->repeat_pict = 1;
 259                         break;
 260                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 261                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 262                         s->repeat_pict = 2;
 263                         break;
 264                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 265                         s->repeat_pict = 3;
 266                         break;
 267                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 268                         s->repeat_pict = 5;
 269                         break;
 270                     default:
 271                         s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 272                         break;
 273                 }
 274             } else {
 275                 s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 276             }
 277
 278             return 0; /* no need to evaluate the rest */
 279         }
 280         buf += h->is_avc ? nalsize : consumed;
 281     }
 282     if (q264)
 283         return 0;
 284     /* didn't find a picture! */
 285     av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 286     return -1;
 287 }
 288
 289 static int h264_parse(AVCodecParserContext *s,
 290                       AVCodecContext *avctx,
 291                       const uint8_t **poutbuf, int *poutbuf_size,
 292                       const uint8_t *buf, int buf_size)
 293 {
 294     H264Context *h = s->priv_data;
 295     ParseContext *pc = &h->s.parse_context;
 296     int next;
 297
 298     if (!h->got_first) {
 299         h->got_first = 1;
 300         if (avctx->extradata_size) {
 301             h->s.avctx = avctx;
 302             // must be done like in decoder, otherwise opening the parser,
 303             // letting it create extradata and then closing and opening again
 304             // will cause has_b_frames to be always set.
 305             // Note that estimate_timings_from_pts does exactly this.
 306             if (!avctx->has_b_frames)
 307                 h->s.low_delay = 1;
 308             ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
 309         }
 310     }
 311
 312     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 313         next= buf_size;
 314     }else{
 315         next= ff_h264_find_frame_end(h, buf, buf_size);
 316
 317         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 318             *poutbuf = NULL;
 319             *poutbuf_size = 0;
 320             return buf_size;
 321         }
 322
 323         if(next<0 && next != END_NOT_FOUND){
 324             av_assert1(pc->last_index + next >= 0 );
 325             ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state
 326         }
 327     }
 328
 329     parse_nal_units(s, avctx, buf, buf_size);
 330
 331     if (h->sei_cpb_removal_delay >= 0) {
 332         s->dts_sync_point    = h->sei_buffering_period_present;
 333         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 334         s->pts_dts_delta     = h->sei_dpb_output_delay;
 335     } else {
 336         s->dts_sync_point    = INT_MIN;
 337         s->dts_ref_dts_delta = INT_MIN;
 338         s->pts_dts_delta     = INT_MIN;
 339     }
 340
 341     if (s->flags & PARSER_FLAG_ONCE) {
 342         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 343     }
 344
 345     *poutbuf = buf;
 346     *poutbuf_size = buf_size;
 347     return next;
 348 }
 349
 350 static int h264_split(AVCodecContext *avctx,
 351                       const uint8_t *buf, int buf_size)
 352 {
 353     int i;
 354     uint32_t state = -1;
 355     int has_sps= 0;
 356
 357     for(i=0; i<=buf_size; i++){
 358         if((state&0xFFFFFF1F) == 0x107)
 359             has_sps=1;
 360 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 361         }*/
 362         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 363             if(has_sps){
 364                 while(i>4 && buf[i-5]==0) i--;
 365                 return i-4;
 366             }
 367         }
 368         if (i<buf_size)
 369             state= (state<<8) | buf[i];
 370     }
 371     return 0;
 372 }
 373
 374 static void close(AVCodecParserContext *s)
 375 {
 376     H264Context *h = s->priv_data;
 377     ParseContext *pc = &h->s.parse_context;
 378
 379     av_free(pc->buffer);
 380     ff_h264_free_context(h);
 381 }
 382
 383 static int init(AVCodecParserContext *s)
 384 {
 385     H264Context *h = s->priv_data;
 386     h->thread_context[0] = h;
 387     h->s.slice_context_count = 1;
 388     return 0;
 389 }
 390
 391 AVCodecParser ff_h264_parser = {
 392     .codec_ids      = { AV_CODEC_ID_H264 },
 393     .priv_data_size = sizeof(H264Context),
 394     .parser_init    = init,
 395     .parser_parse   = h264_parse,
 396     .parser_close   = close,
 397     .split          = h264_split,
 398 };