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 //printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]);
  43 //    mb_addr= pc->mb_addr - 1;
  44     state= pc->state;
  45     if(state>13)
  46         state= 7;
  47
  48     if(h->is_avc && !h->nal_length_size)
  49         av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
  50
  51     for(i=0; i<buf_size; i++){
  52         if(i >= next_avc) {
  53             int nalsize = 0;
  54             i = next_avc;
  55             for(j = 0; j < h->nal_length_size; j++)
  56                 nalsize = (nalsize << 8) | buf[i++];
  57             if(nalsize <= 0 || nalsize > buf_size - i){
  58                 av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
  59                 return buf_size;
  60             }
  61             next_avc= i + nalsize;
  62             state= 5;
  63         }
  64
  65         if(state==7){
  66 #if HAVE_FAST_UNALIGNED
  67         /* we check i<buf_size instead of i+3/7 because its simpler
  68          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  69          */
  70 #    if HAVE_FAST_64BIT
  71             while(i<next_avc && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  72                 i+=8;
  73 #    else
  74             while(i<next_avc && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  75                 i+=4;
  76 #    endif
  77 #endif
  78             for(; i<next_avc; i++){
  79                 if(!buf[i]){
  80                     state=2;
  81                     break;
  82                 }
  83             }
  84         }else if(state<=2){
  85             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  86             else if(buf[i]) state = 7;
  87             else            state>>=1; //2->1, 1->0, 0->0
  88         }else if(state<=5){
  89             int v= buf[i] & 0x1F;
  90             if(v==6 || v==7 || v==8 || v==9){
  91                 if(pc->frame_start_found){
  92                     i++;
  93                     goto found;
  94                 }
  95             }else if(v==1 || v==2 || v==5){
  96                 state+=8;
  97                 continue;
  98             }
  99             state= 7;
 100         }else{
 101             h->parse_history[h->parse_history_count++]= buf[i];
 102             if(h->parse_history_count>3){
 103                 unsigned int mb, last_mb= h->parse_last_mb;
 104                 GetBitContext gb;
 105
 106                 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
 107                 h->parse_history_count=0;
 108                 mb= get_ue_golomb_long(&gb);
 109                 last_mb= h->parse_last_mb;
 110                 h->parse_last_mb= mb;
 111                 if(pc->frame_start_found){
 112                     if(mb <= last_mb)
 113                         goto found;
 114                 }else
 115                     pc->frame_start_found = 1;
 116                 state= 7;
 117             }
 118         }
 119     }
 120     pc->state= state;
 121     if(h->is_avc)
 122         return next_avc;
 123     return END_NOT_FOUND;
 124
 125 found:
 126     pc->state=7;
 127     pc->frame_start_found= 0;
 128     if(h->is_avc)
 129         return next_avc;
 130     return i-(state&5) - 3*(state>7);
 131 }
 132
 133 /**
 134  * Parse NAL units of found picture and decode some basic information.
 135  *
 136  * @param s parser context.
 137  * @param avctx codec context.
 138  * @param buf buffer with field/frame data.
 139  * @param buf_size size of the buffer.
 140  */
 141 static inline int parse_nal_units(AVCodecParserContext *s,
 142                                   AVCodecContext *avctx,
 143                                   const uint8_t *buf, int buf_size)
 144 {
 145     H264Context *h = s->priv_data;
 146     const uint8_t *buf_end = buf + buf_size;
 147     unsigned int pps_id;
 148     unsigned int slice_type;
 149     int state = -1;
 150     const uint8_t *ptr;
 151     int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
 152
 153     /* set some sane default values */
 154     s->pict_type = AV_PICTURE_TYPE_I;
 155     s->key_frame = 0;
 156
 157     h->s.avctx= avctx;
 158     h->sei_recovery_frame_cnt = -1;
 159     h->sei_dpb_output_delay         =  0;
 160     h->sei_cpb_removal_delay        = -1;
 161     h->sei_buffering_period_present =  0;
 162
 163     if (!buf_size)
 164         return 0;
 165
 166     for(;;) {
 167         int src_length, dst_length, consumed, nalsize = 0;
 168         if (h->is_avc) {
 169             int i;
 170             if (h->nal_length_size >= buf_end - buf) break;
 171             nalsize = 0;
 172             for (i = 0; i < h->nal_length_size; i++)
 173                 nalsize = (nalsize << 8) | *buf++;
 174             if (nalsize <= 0 || nalsize > buf_end - buf) {
 175                 av_log(h->s.avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
 176                 break;
 177             }
 178             src_length = nalsize;
 179         } else {
 180         buf = avpriv_mpv_find_start_code(buf, buf_end, &state);
 181         if(buf >= buf_end)
 182             break;
 183         --buf;
 184         src_length = buf_end - buf;
 185         }
 186         switch (state & 0x1f) {
 187         case NAL_SLICE:
 188         case NAL_IDR_SLICE:
 189             // Do not walk the whole buffer just to decode slice header
 190             if (src_length > 20)
 191                 src_length = 20;
 192             break;
 193         }
 194         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 195         if (ptr==NULL || dst_length < 0)
 196             break;
 197
 198         init_get_bits(&h->s.gb, ptr, 8*dst_length);
 199         switch(h->nal_unit_type) {
 200         case NAL_SPS:
 201             ff_h264_decode_seq_parameter_set(h);
 202             break;
 203         case NAL_PPS:
 204             ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
 205             break;
 206         case NAL_SEI:
 207             ff_h264_decode_sei(h);
 208             break;
 209         case NAL_IDR_SLICE:
 210             s->key_frame = 1;
 211             /* fall through */
 212         case NAL_SLICE:
 213             get_ue_golomb_long(&h->s.gb);  // skip first_mb_in_slice
 214             slice_type = get_ue_golomb_31(&h->s.gb);
 215             s->pict_type = golomb_to_pict_type[slice_type % 5];
 216             if (h->sei_recovery_frame_cnt >= 0) {
 217                 /* key frame, since recovery_frame_cnt is set */
 218                 s->key_frame = 1;
 219             }
 220             pps_id= get_ue_golomb(&h->s.gb);
 221             if(pps_id>=MAX_PPS_COUNT) {
 222                 av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
 223                 return -1;
 224             }
 225             if(!h->pps_buffers[pps_id]) {
 226                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 227                 return -1;
 228             }
 229             h->pps= *h->pps_buffers[pps_id];
 230             if(!h->sps_buffers[h->pps.sps_id]) {
 231                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 232                 return -1;
 233             }
 234             h->sps = *h->sps_buffers[h->pps.sps_id];
 235             h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
 236
 237             avctx->profile = ff_h264_get_profile(&h->sps);
 238             avctx->level   = h->sps.level_idc;
 239
 240             if(h->sps.frame_mbs_only_flag){
 241                 h->s.picture_structure= PICT_FRAME;
 242             }else{
 243                 if(get_bits1(&h->s.gb)) { //field_pic_flag
 244                     h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
 245                 } else {
 246                     h->s.picture_structure= PICT_FRAME;
 247                 }
 248             }
 249
 250             if(h->sps.pic_struct_present_flag) {
 251                 switch (h->sei_pic_struct) {
 252                     case SEI_PIC_STRUCT_TOP_FIELD:
 253                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 254                         s->repeat_pict = 0;
 255                         break;
 256                     case SEI_PIC_STRUCT_FRAME:
 257                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 258                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 259                         s->repeat_pict = 1;
 260                         break;
 261                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 262                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 263                         s->repeat_pict = 2;
 264                         break;
 265                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 266                         s->repeat_pict = 3;
 267                         break;
 268                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 269                         s->repeat_pict = 5;
 270                         break;
 271                     default:
 272                         s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 273                         break;
 274                 }
 275             } else {
 276                 s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 277             }
 278
 279             return 0; /* no need to evaluate the rest */
 280         }
 281         buf += h->is_avc ? nalsize : consumed;
 282     }
 283     if (q264)
 284         return 0;
 285     /* didn't find a picture! */
 286     av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 287     return -1;
 288 }
 289
 290 static int h264_parse(AVCodecParserContext *s,
 291                       AVCodecContext *avctx,
 292                       const uint8_t **poutbuf, int *poutbuf_size,
 293                       const uint8_t *buf, int buf_size)
 294 {
 295     H264Context *h = s->priv_data;
 296     ParseContext *pc = &h->s.parse_context;
 297     int next;
 298
 299     if (!h->got_first) {
 300         h->got_first = 1;
 301         if (avctx->extradata_size) {
 302             h->s.avctx = avctx;
 303             // must be done like in decoder, otherwise opening the parser,
 304             // letting it create extradata and then closing and opening again
 305             // will cause has_b_frames to be always set.
 306             // Note that estimate_timings_from_pts does exactly this.
 307             if (!avctx->has_b_frames)
 308                 h->s.low_delay = 1;
 309             ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
 310         }
 311     }
 312
 313     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 314         next= buf_size;
 315     }else{
 316         next= ff_h264_find_frame_end(h, buf, buf_size);
 317
 318         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 319             *poutbuf = NULL;
 320             *poutbuf_size = 0;
 321             return buf_size;
 322         }
 323
 324         if(next<0 && next != END_NOT_FOUND){
 325             av_assert1(pc->last_index + next >= 0 );
 326             ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state
 327         }
 328     }
 329
 330     parse_nal_units(s, avctx, buf, buf_size);
 331
 332     if (h->sei_cpb_removal_delay >= 0) {
 333         s->dts_sync_point    = h->sei_buffering_period_present;
 334         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 335         s->pts_dts_delta     = h->sei_dpb_output_delay;
 336     } else {
 337         s->dts_sync_point    = INT_MIN;
 338         s->dts_ref_dts_delta = INT_MIN;
 339         s->pts_dts_delta     = INT_MIN;
 340     }
 341
 342     if (s->flags & PARSER_FLAG_ONCE) {
 343         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 344     }
 345
 346     *poutbuf = buf;
 347     *poutbuf_size = buf_size;
 348     return next;
 349 }
 350
 351 static int h264_split(AVCodecContext *avctx,
 352                       const uint8_t *buf, int buf_size)
 353 {
 354     int i;
 355     uint32_t state = -1;
 356     int has_sps= 0;
 357
 358     for(i=0; i<=buf_size; i++){
 359         if((state&0xFFFFFF1F) == 0x107)
 360             has_sps=1;
 361 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 362         }*/
 363         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 364             if(has_sps){
 365                 while(i>4 && buf[i-5]==0) i--;
 366                 return i-4;
 367             }
 368         }
 369         if (i<buf_size)
 370             state= (state<<8) | buf[i];
 371     }
 372     return 0;
 373 }
 374
 375 static void close(AVCodecParserContext *s)
 376 {
 377     H264Context *h = s->priv_data;
 378     ParseContext *pc = &h->s.parse_context;
 379
 380     av_free(pc->buffer);
 381     ff_h264_free_context(h);
 382 }
 383
 384 static int init(AVCodecParserContext *s)
 385 {
 386     H264Context *h = s->priv_data;
 387     h->thread_context[0] = h;
 388     h->s.slice_context_count = 1;
 389     return 0;
 390 }
 391
 392 AVCodecParser ff_h264_parser = {
 393     .codec_ids      = { AV_CODEC_ID_H264 },
 394     .priv_data_size = sizeof(H264Context),
 395     .parser_init    = init,
 396     .parser_parse   = h264_parse,
 397     .parser_close   = close,
 398     .split          = h264_split,
 399 };