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 "libavutil/attributes.h"
  31 #include "parser.h"
  32 #include "h264data.h"
  33 #include "golomb.h"
  34 #include "internal.h"
  35
  36
  37 static int h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_size)
  38 {
  39     int i, j;
  40     uint32_t state;
  41     ParseContext *pc = &h->parse_context;
  42     int next_avc= h->is_avc ? 0 : buf_size;
  43
  44 //    mb_addr= pc->mb_addr - 1;
  45     state= pc->state;
  46     if(state>13)
  47         state= 7;
  48
  49     if(h->is_avc && !h->nal_length_size)
  50         av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
  51
  52     for(i=0; i<buf_size; i++){
  53         if(i >= next_avc) {
  54             int nalsize = 0;
  55             i = next_avc;
  56             for(j = 0; j < h->nal_length_size; j++)
  57                 nalsize = (nalsize << 8) | buf[i++];
  58             if(nalsize <= 0 || nalsize > buf_size - i){
  59                 av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
  60                 return buf_size;
  61             }
  62             next_avc= i + nalsize;
  63             state= 5;
  64         }
  65
  66         if(state==7){
  67 #if HAVE_FAST_UNALIGNED
  68         /* we check i<buf_size instead of i+3/7 because its simpler
  69          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  70          */
  71 #    if HAVE_FAST_64BIT
  72             while(i<next_avc && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  73                 i+=8;
  74 #    else
  75             while(i<next_avc && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  76                 i+=4;
  77 #    endif
  78 #endif
  79             for(; i<next_avc; i++){
  80                 if(!buf[i]){
  81                     state=2;
  82                     break;
  83                 }
  84             }
  85         }else if(state<=2){
  86             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  87             else if(buf[i]) state = 7;
  88             else            state>>=1; //2->1, 1->0, 0->0
  89         }else if(state<=5){
  90             int v= buf[i] & 0x1F;
  91             if(v==6 || v==7 || v==8 || v==9){
  92                 if(pc->frame_start_found){
  93                     i++;
  94                     goto found;
  95                 }
  96             }else if(v==1 || v==2 || v==5){
  97                 state+=8;
  98                 continue;
  99             }
 100             state= 7;
 101         }else{
 102             h->parse_history[h->parse_history_count++]= buf[i];
 103             if(h->parse_history_count>3){
 104                 unsigned int mb, last_mb= h->parse_last_mb;
 105                 GetBitContext gb;
 106
 107                 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
 108                 h->parse_history_count=0;
 109                 mb= get_ue_golomb_long(&gb);
 110                 last_mb= h->parse_last_mb;
 111                 h->parse_last_mb= mb;
 112                 if(pc->frame_start_found){
 113                     if(mb <= last_mb)
 114                         goto found;
 115                 }else
 116                     pc->frame_start_found = 1;
 117                 state= 7;
 118             }
 119         }
 120     }
 121     pc->state= state;
 122     if(h->is_avc)
 123         return next_avc;
 124     return END_NOT_FOUND;
 125
 126 found:
 127     pc->state=7;
 128     pc->frame_start_found= 0;
 129     if(h->is_avc)
 130         return next_avc;
 131     return i-(state&5) - 3*(state>7);
 132 }
 133
 134 /**
 135  * Parse NAL units of found picture and decode some basic information.
 136  *
 137  * @param s parser context.
 138  * @param avctx codec context.
 139  * @param buf buffer with field/frame data.
 140  * @param buf_size size of the buffer.
 141  */
 142 static inline int parse_nal_units(AVCodecParserContext *s,
 143                                   AVCodecContext *avctx,
 144                                   const uint8_t *buf, int buf_size)
 145 {
 146     H264Context *h = s->priv_data;
 147     const uint8_t *buf_end = buf + buf_size;
 148     unsigned int pps_id;
 149     unsigned int slice_type;
 150     int state = -1;
 151     const uint8_t *ptr;
 152     int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
 153
 154     /* set some sane default values */
 155     s->pict_type = AV_PICTURE_TYPE_I;
 156     s->key_frame = 0;
 157
 158     h->avctx= avctx;
 159     h->sei_recovery_frame_cnt = -1;
 160     h->sei_dpb_output_delay         =  0;
 161     h->sei_cpb_removal_delay        = -1;
 162     h->sei_buffering_period_present =  0;
 163
 164     if (!buf_size)
 165         return 0;
 166
 167     for(;;) {
 168         int src_length, dst_length, consumed, nalsize = 0;
 169         if (h->is_avc) {
 170             int i;
 171             if (h->nal_length_size >= buf_end - buf) break;
 172             nalsize = 0;
 173             for (i = 0; i < h->nal_length_size; i++)
 174                 nalsize = (nalsize << 8) | *buf++;
 175             if (nalsize <= 0 || nalsize > buf_end - buf) {
 176                 av_log(h->avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize);
 177                 break;
 178             }
 179             src_length = nalsize;
 180         } else {
 181         buf = avpriv_find_start_code(buf, buf_end, &state);
 182         if(buf >= buf_end)
 183             break;
 184         --buf;
 185         src_length = buf_end - buf;
 186         }
 187         switch (state & 0x1f) {
 188         case NAL_SLICE:
 189         case NAL_IDR_SLICE:
 190             // Do not walk the whole buffer just to decode slice header
 191             if (src_length > 20)
 192                 src_length = 20;
 193             break;
 194         }
 195         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 196         if (ptr==NULL || dst_length < 0)
 197             break;
 198
 199         init_get_bits(&h->gb, ptr, 8*dst_length);
 200         switch(h->nal_unit_type) {
 201         case NAL_SPS:
 202             ff_h264_decode_seq_parameter_set(h);
 203             break;
 204         case NAL_PPS:
 205             ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
 206             break;
 207         case NAL_SEI:
 208             ff_h264_decode_sei(h);
 209             break;
 210         case NAL_IDR_SLICE:
 211             s->key_frame = 1;
 212             /* fall through */
 213         case NAL_SLICE:
 214             get_ue_golomb_long(&h->gb);  // skip first_mb_in_slice
 215             slice_type = get_ue_golomb_31(&h->gb);
 216             s->pict_type = golomb_to_pict_type[slice_type % 5];
 217             if (h->sei_recovery_frame_cnt >= 0) {
 218                 /* key frame, since recovery_frame_cnt is set */
 219                 s->key_frame = 1;
 220             }
 221             pps_id= get_ue_golomb(&h->gb);
 222             if(pps_id>=MAX_PPS_COUNT) {
 223                 av_log(h->avctx, AV_LOG_ERROR, "pps_id out of range\n");
 224                 return -1;
 225             }
 226             if(!h->pps_buffers[pps_id]) {
 227                 av_log(h->avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 228                 return -1;
 229             }
 230             h->pps= *h->pps_buffers[pps_id];
 231             if(!h->sps_buffers[h->pps.sps_id]) {
 232                 av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 233                 return -1;
 234             }
 235             h->sps = *h->sps_buffers[h->pps.sps_id];
 236             h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
 237
 238             avctx->profile = ff_h264_get_profile(&h->sps);
 239             avctx->level   = h->sps.level_idc;
 240
 241             if(h->sps.frame_mbs_only_flag){
 242                 h->picture_structure= PICT_FRAME;
 243             }else{
 244                 if(get_bits1(&h->gb)) { //field_pic_flag
 245                     h->picture_structure= PICT_TOP_FIELD + get_bits1(&h->gb); //bottom_field_flag
 246                 } else {
 247                     h->picture_structure= PICT_FRAME;
 248                 }
 249             }
 250
 251             if(h->sps.pic_struct_present_flag) {
 252                 switch (h->sei_pic_struct) {
 253                     case SEI_PIC_STRUCT_TOP_FIELD:
 254                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 255                         s->repeat_pict = 0;
 256                         break;
 257                     case SEI_PIC_STRUCT_FRAME:
 258                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 259                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 260                         s->repeat_pict = 1;
 261                         break;
 262                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 263                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 264                         s->repeat_pict = 2;
 265                         break;
 266                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 267                         s->repeat_pict = 3;
 268                         break;
 269                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 270                         s->repeat_pict = 5;
 271                         break;
 272                     default:
 273                         s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 274                         break;
 275                 }
 276             } else {
 277                 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 278             }
 279
 280             return 0; /* no need to evaluate the rest */
 281         }
 282         buf += h->is_avc ? nalsize : consumed;
 283     }
 284     if (q264)
 285         return 0;
 286     /* didn't find a picture! */
 287     av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 288     return -1;
 289 }
 290
 291 static int h264_parse(AVCodecParserContext *s,
 292                       AVCodecContext *avctx,
 293                       const uint8_t **poutbuf, int *poutbuf_size,
 294                       const uint8_t *buf, int buf_size)
 295 {
 296     H264Context *h = s->priv_data;
 297     ParseContext *pc = &h->parse_context;
 298     int next;
 299
 300     if (!h->got_first) {
 301         h->got_first = 1;
 302         if (avctx->extradata_size) {
 303             h->avctx = avctx;
 304             // must be done like in decoder, otherwise opening the parser,
 305             // letting it create extradata and then closing and opening again
 306             // will cause has_b_frames to be always set.
 307             // Note that estimate_timings_from_pts does exactly this.
 308             if (!avctx->has_b_frames)
 309                 h->low_delay = 1;
 310             ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
 311         }
 312     }
 313
 314     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 315         next= buf_size;
 316     }else{
 317         next = h264_find_frame_end(h, buf, buf_size);
 318
 319         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 320             *poutbuf = NULL;
 321             *poutbuf_size = 0;
 322             return buf_size;
 323         }
 324
 325         if(next<0 && next != END_NOT_FOUND){
 326             av_assert1(pc->last_index + next >= 0 );
 327             h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
 328         }
 329     }
 330
 331     parse_nal_units(s, avctx, buf, buf_size);
 332
 333     if (h->sei_cpb_removal_delay >= 0) {
 334         s->dts_sync_point    = h->sei_buffering_period_present;
 335         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 336         s->pts_dts_delta     = h->sei_dpb_output_delay;
 337     } else {
 338         s->dts_sync_point    = INT_MIN;
 339         s->dts_ref_dts_delta = INT_MIN;
 340         s->pts_dts_delta     = INT_MIN;
 341     }
 342
 343     if (s->flags & PARSER_FLAG_ONCE) {
 344         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 345     }
 346
 347     *poutbuf = buf;
 348     *poutbuf_size = buf_size;
 349     return next;
 350 }
 351
 352 static int h264_split(AVCodecContext *avctx,
 353                       const uint8_t *buf, int buf_size)
 354 {
 355     int i;
 356     uint32_t state = -1;
 357     int has_sps= 0;
 358
 359     for(i=0; i<=buf_size; i++){
 360         if((state&0xFFFFFF1F) == 0x107)
 361             has_sps=1;
 362 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 363         }*/
 364         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 365             if(has_sps){
 366                 while(i>4 && buf[i-5]==0) i--;
 367                 return i-4;
 368             }
 369         }
 370         if (i<buf_size)
 371             state= (state<<8) | buf[i];
 372     }
 373     return 0;
 374 }
 375
 376 static void close(AVCodecParserContext *s)
 377 {
 378     H264Context *h = s->priv_data;
 379     ParseContext *pc = &h->parse_context;
 380
 381     av_free(pc->buffer);
 382     ff_h264_free_context(h);
 383 }
 384
 385 static av_cold int init(AVCodecParserContext *s)
 386 {
 387     H264Context *h = s->priv_data;
 388     h->thread_context[0] = h;
 389     h->slice_context_count = 1;
 390     return 0;
 391 }
 392
 393 AVCodecParser ff_h264_parser = {
 394     .codec_ids      = { AV_CODEC_ID_H264 },
 395     .priv_data_size = sizeof(H264Context),
 396     .parser_init    = init,
 397     .parser_parse   = h264_parse,
 398     .parser_close   = close,
 399     .split          = h264_split,
 400 };