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 #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, 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
 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;
 167         buf = avpriv_mpv_find_start_code(buf, buf_end, &state);
 168         if(buf >= buf_end)
 169             break;
 170         --buf;
 171         src_length = buf_end - buf;
 172         switch (state & 0x1f) {
 173         case NAL_SLICE:
 174         case NAL_IDR_SLICE:
 175             // Do not walk the whole buffer just to decode slice header
 176             if (src_length > 20)
 177                 src_length = 20;
 178             break;
 179         }
 180         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 181         if (ptr==NULL || dst_length < 0)
 182             break;
 183
 184         init_get_bits(&h->s.gb, ptr, 8*dst_length);
 185         switch(h->nal_unit_type) {
 186         case NAL_SPS:
 187             ff_h264_decode_seq_parameter_set(h);
 188             break;
 189         case NAL_PPS:
 190             ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
 191             break;
 192         case NAL_SEI:
 193             ff_h264_decode_sei(h);
 194             break;
 195         case NAL_IDR_SLICE:
 196             s->key_frame = 1;
 197             /* fall through */
 198         case NAL_SLICE:
 199             get_ue_golomb_long(&h->s.gb);  // skip first_mb_in_slice
 200             slice_type = get_ue_golomb_31(&h->s.gb);
 201             s->pict_type = golomb_to_pict_type[slice_type % 5];
 202             if (h->sei_recovery_frame_cnt >= 0) {
 203                 /* key frame, since recovery_frame_cnt is set */
 204                 s->key_frame = 1;
 205             }
 206             pps_id= get_ue_golomb(&h->s.gb);
 207             if(pps_id>=MAX_PPS_COUNT) {
 208                 av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
 209                 return -1;
 210             }
 211             if(!h->pps_buffers[pps_id]) {
 212                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 213                 return -1;
 214             }
 215             h->pps= *h->pps_buffers[pps_id];
 216             if(!h->sps_buffers[h->pps.sps_id]) {
 217                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 218                 return -1;
 219             }
 220             h->sps = *h->sps_buffers[h->pps.sps_id];
 221             h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
 222
 223             avctx->profile = ff_h264_get_profile(&h->sps);
 224             avctx->level   = h->sps.level_idc;
 225
 226             if(h->sps.frame_mbs_only_flag){
 227                 h->s.picture_structure= PICT_FRAME;
 228             }else{
 229                 if(get_bits1(&h->s.gb)) { //field_pic_flag
 230                     h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
 231                 } else {
 232                     h->s.picture_structure= PICT_FRAME;
 233                 }
 234             }
 235
 236             if(h->sps.pic_struct_present_flag) {
 237                 switch (h->sei_pic_struct) {
 238                     case SEI_PIC_STRUCT_TOP_FIELD:
 239                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 240                         s->repeat_pict = 0;
 241                         break;
 242                     case SEI_PIC_STRUCT_FRAME:
 243                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 244                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 245                         s->repeat_pict = 1;
 246                         break;
 247                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 248                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 249                         s->repeat_pict = 2;
 250                         break;
 251                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 252                         s->repeat_pict = 3;
 253                         break;
 254                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 255                         s->repeat_pict = 5;
 256                         break;
 257                     default:
 258                         s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 259                         break;
 260                 }
 261             } else {
 262                 s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 263             }
 264
 265             return 0; /* no need to evaluate the rest */
 266         }
 267         buf += consumed;
 268     }
 269     /* didn't find a picture! */
 270     av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 271     return -1;
 272 }
 273
 274 static int h264_parse(AVCodecParserContext *s,
 275                       AVCodecContext *avctx,
 276                       const uint8_t **poutbuf, int *poutbuf_size,
 277                       const uint8_t *buf, int buf_size)
 278 {
 279     H264Context *h = s->priv_data;
 280     ParseContext *pc = &h->s.parse_context;
 281     int next;
 282
 283     if (!h->got_first) {
 284         h->got_first = 1;
 285         if (avctx->extradata_size) {
 286             h->s.avctx = avctx;
 287             // must be done like in decoder, otherwise opening the parser,
 288             // letting it create extradata and then closing and opening again
 289             // will cause has_b_frames to be always set.
 290             // Note that estimate_timings_from_pts does exactly this.
 291             if (!avctx->has_b_frames)
 292                 h->s.low_delay = 1;
 293             ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
 294         }
 295     }
 296
 297     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 298         next= buf_size;
 299     }else{
 300         next= ff_h264_find_frame_end(h, buf, buf_size);
 301
 302         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 303             *poutbuf = NULL;
 304             *poutbuf_size = 0;
 305             return buf_size;
 306         }
 307
 308         if(next<0 && next != END_NOT_FOUND){
 309             assert(pc->last_index + next >= 0 );
 310             ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state
 311         }
 312     }
 313
 314     if(!h->is_avc){
 315     parse_nal_units(s, avctx, buf, buf_size);
 316
 317     if (h->sei_cpb_removal_delay >= 0) {
 318         s->dts_sync_point    = h->sei_buffering_period_present;
 319         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 320         s->pts_dts_delta     = h->sei_dpb_output_delay;
 321     } else {
 322         s->dts_sync_point    = INT_MIN;
 323         s->dts_ref_dts_delta = INT_MIN;
 324         s->pts_dts_delta     = INT_MIN;
 325     }
 326
 327     if (s->flags & PARSER_FLAG_ONCE) {
 328         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 329     }
 330     }
 331
 332     *poutbuf = buf;
 333     *poutbuf_size = buf_size;
 334     return next;
 335 }
 336
 337 static int h264_split(AVCodecContext *avctx,
 338                       const uint8_t *buf, int buf_size)
 339 {
 340     int i;
 341     uint32_t state = -1;
 342     int has_sps= 0;
 343
 344     for(i=0; i<=buf_size; i++){
 345         if((state&0xFFFFFF1F) == 0x107)
 346             has_sps=1;
 347 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 348         }*/
 349         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 350             if(has_sps){
 351                 while(i>4 && buf[i-5]==0) i--;
 352                 return i-4;
 353             }
 354         }
 355         if (i<buf_size)
 356             state= (state<<8) | buf[i];
 357     }
 358     return 0;
 359 }
 360
 361 static void close(AVCodecParserContext *s)
 362 {
 363     H264Context *h = s->priv_data;
 364     ParseContext *pc = &h->s.parse_context;
 365
 366     av_free(pc->buffer);
 367     ff_h264_free_context(h);
 368 }
 369
 370 static int init(AVCodecParserContext *s)
 371 {
 372     H264Context *h = s->priv_data;
 373     h->thread_context[0] = h;
 374     return 0;
 375 }
 376
 377 AVCodecParser ff_h264_parser = {
 378     .codec_ids      = { CODEC_ID_H264 },
 379     .priv_data_size = sizeof(H264Context),
 380     .parser_init    = init,
 381     .parser_parse   = h264_parse,
 382     .parser_close   = close,
 383     .split          = h264_split,
 384 };