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 #include <assert.h>
  35
  36
  37 static int ff_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->s.parse_context);
  42     int next_avc= h->is_avc ? 0 : buf_size;
  43
  44 //printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]);
  45 //    mb_addr= pc->mb_addr - 1;
  46     state= pc->state;
  47     if(state>13)
  48         state= 7;
  49
  50     if(h->is_avc && !h->nal_length_size)
  51         av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
  52
  53     for(i=0; i<buf_size; i++){
  54         if(i >= next_avc) {
  55             int nalsize = 0;
  56             i = next_avc;
  57             for(j = 0; j < h->nal_length_size; j++)
  58                 nalsize = (nalsize << 8) | buf[i++];
  59             if(nalsize <= 0 || nalsize > buf_size - i){
  60                 av_log(h->s.avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
  61                 return buf_size;
  62             }
  63             next_avc= i + nalsize;
  64             state= 5;
  65         }
  66
  67         if(state==7){
  68 #if HAVE_FAST_UNALIGNED
  69         /* we check i<buf_size instead of i+3/7 because its simpler
  70          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  71          */
  72 #    if HAVE_FAST_64BIT
  73             while(i<next_avc && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  74                 i+=8;
  75 #    else
  76             while(i<next_avc && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  77                 i+=4;
  78 #    endif
  79 #endif
  80             for(; i<next_avc; i++){
  81                 if(!buf[i]){
  82                     state=2;
  83                     break;
  84                 }
  85             }
  86         }else if(state<=2){
  87             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  88             else if(buf[i]) state = 7;
  89             else            state>>=1; //2->1, 1->0, 0->0
  90         }else if(state<=5){
  91             int v= buf[i] & 0x1F;
  92             if(v==6 || v==7 || v==8 || v==9){
  93                 if(pc->frame_start_found){
  94                     i++;
  95                     goto found;
  96                 }
  97             }else if(v==1 || v==2 || v==5){
  98                 state+=8;
  99                 continue;
 100             }
 101             state= 7;
 102         }else{
 103             h->parse_history[h->parse_history_count++]= buf[i];
 104             if(h->parse_history_count>3){
 105                 unsigned int mb, last_mb= h->parse_last_mb;
 106                 GetBitContext gb;
 107
 108                 init_get_bits(&gb, h->parse_history, 8*h->parse_history_count);
 109                 h->parse_history_count=0;
 110                 mb= get_ue_golomb_long(&gb);
 111                 last_mb= h->parse_last_mb;
 112                 h->parse_last_mb= mb;
 113                 if(pc->frame_start_found){
 114                     if(mb <= last_mb)
 115                         goto found;
 116                 }else
 117                     pc->frame_start_found = 1;
 118                 state= 7;
 119             }
 120         }
 121     }
 122     pc->state= state;
 123     if(h->is_avc)
 124         return next_avc;
 125     return END_NOT_FOUND;
 126
 127 found:
 128     pc->state=7;
 129     pc->frame_start_found= 0;
 130     if(h->is_avc)
 131         return next_avc;
 132     return i-(state&5) - 3*(state>7);
 133 }
 134
 135 /**
 136  * Parse NAL units of found picture and decode some basic information.
 137  *
 138  * @param s parser context.
 139  * @param avctx codec context.
 140  * @param buf buffer with field/frame data.
 141  * @param buf_size size of the buffer.
 142  */
 143 static inline int parse_nal_units(AVCodecParserContext *s,
 144                                   AVCodecContext *avctx,
 145                                   const uint8_t *buf, int buf_size)
 146 {
 147     H264Context *h = s->priv_data;
 148     const uint8_t *buf_end = buf + buf_size;
 149     unsigned int pps_id;
 150     unsigned int slice_type;
 151     int state = -1;
 152     const uint8_t *ptr;
 153     int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
 154
 155     /* set some sane default values */
 156     s->pict_type = AV_PICTURE_TYPE_I;
 157     s->key_frame = 0;
 158
 159     h->s.avctx= avctx;
 160     h->sei_recovery_frame_cnt = -1;
 161     h->sei_dpb_output_delay         =  0;
 162     h->sei_cpb_removal_delay        = -1;
 163     h->sei_buffering_period_present =  0;
 164
 165     if (!buf_size)
 166         return 0;
 167
 168     for(;;) {
 169         int src_length, dst_length, consumed;
 170         buf = avpriv_mpv_find_start_code(buf, buf_end, &state);
 171         if(buf >= buf_end)
 172             break;
 173         --buf;
 174         src_length = buf_end - buf;
 175         switch (state & 0x1f) {
 176         case NAL_SLICE:
 177         case NAL_IDR_SLICE:
 178             // Do not walk the whole buffer just to decode slice header
 179             if (src_length > 20)
 180                 src_length = 20;
 181             break;
 182         }
 183         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 184         if (ptr==NULL || dst_length < 0)
 185             break;
 186
 187         init_get_bits(&h->s.gb, ptr, 8*dst_length);
 188         switch(h->nal_unit_type) {
 189         case NAL_SPS:
 190             ff_h264_decode_seq_parameter_set(h);
 191             break;
 192         case NAL_PPS:
 193             ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits);
 194             break;
 195         case NAL_SEI:
 196             ff_h264_decode_sei(h);
 197             break;
 198         case NAL_IDR_SLICE:
 199             s->key_frame = 1;
 200             /* fall through */
 201         case NAL_SLICE:
 202             get_ue_golomb_long(&h->s.gb);  // skip first_mb_in_slice
 203             slice_type = get_ue_golomb_31(&h->s.gb);
 204             s->pict_type = golomb_to_pict_type[slice_type % 5];
 205             if (h->sei_recovery_frame_cnt >= 0) {
 206                 /* key frame, since recovery_frame_cnt is set */
 207                 s->key_frame = 1;
 208             }
 209             pps_id= get_ue_golomb(&h->s.gb);
 210             if(pps_id>=MAX_PPS_COUNT) {
 211                 av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
 212                 return -1;
 213             }
 214             if(!h->pps_buffers[pps_id]) {
 215                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 216                 return -1;
 217             }
 218             h->pps= *h->pps_buffers[pps_id];
 219             if(!h->sps_buffers[h->pps.sps_id]) {
 220                 av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 221                 return -1;
 222             }
 223             h->sps = *h->sps_buffers[h->pps.sps_id];
 224             h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num);
 225
 226             avctx->profile = ff_h264_get_profile(&h->sps);
 227             avctx->level   = h->sps.level_idc;
 228
 229             if(h->sps.frame_mbs_only_flag){
 230                 h->s.picture_structure= PICT_FRAME;
 231             }else{
 232                 if(get_bits1(&h->s.gb)) { //field_pic_flag
 233                     h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag
 234                 } else {
 235                     h->s.picture_structure= PICT_FRAME;
 236                 }
 237             }
 238
 239             if(h->sps.pic_struct_present_flag) {
 240                 switch (h->sei_pic_struct) {
 241                     case SEI_PIC_STRUCT_TOP_FIELD:
 242                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 243                         s->repeat_pict = 0;
 244                         break;
 245                     case SEI_PIC_STRUCT_FRAME:
 246                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 247                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 248                         s->repeat_pict = 1;
 249                         break;
 250                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 251                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 252                         s->repeat_pict = 2;
 253                         break;
 254                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 255                         s->repeat_pict = 3;
 256                         break;
 257                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 258                         s->repeat_pict = 5;
 259                         break;
 260                     default:
 261                         s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 262                         break;
 263                 }
 264             } else {
 265                 s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0;
 266             }
 267
 268             return 0; /* no need to evaluate the rest */
 269         }
 270         buf += consumed;
 271     }
 272     if (q264)
 273         return 0;
 274     /* didn't find a picture! */
 275     av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 276     return -1;
 277 }
 278
 279 static int h264_parse(AVCodecParserContext *s,
 280                       AVCodecContext *avctx,
 281                       const uint8_t **poutbuf, int *poutbuf_size,
 282                       const uint8_t *buf, int buf_size)
 283 {
 284     H264Context *h = s->priv_data;
 285     ParseContext *pc = &h->s.parse_context;
 286     int next;
 287
 288     if (!h->got_first) {
 289         h->got_first = 1;
 290         if (avctx->extradata_size) {
 291             h->s.avctx = avctx;
 292             // must be done like in decoder, otherwise opening the parser,
 293             // letting it create extradata and then closing and opening again
 294             // will cause has_b_frames to be always set.
 295             // Note that estimate_timings_from_pts does exactly this.
 296             if (!avctx->has_b_frames)
 297                 h->s.low_delay = 1;
 298             ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size);
 299         }
 300     }
 301
 302     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 303         next= buf_size;
 304     }else{
 305         next= ff_h264_find_frame_end(h, buf, buf_size);
 306
 307         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 308             *poutbuf = NULL;
 309             *poutbuf_size = 0;
 310             return buf_size;
 311         }
 312
 313         if(next<0 && next != END_NOT_FOUND){
 314             assert(pc->last_index + next >= 0 );
 315             ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state
 316         }
 317     }
 318
 319     if(!h->is_avc){
 320     parse_nal_units(s, avctx, buf, buf_size);
 321
 322     if (h->sei_cpb_removal_delay >= 0) {
 323         s->dts_sync_point    = h->sei_buffering_period_present;
 324         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 325         s->pts_dts_delta     = h->sei_dpb_output_delay;
 326     } else {
 327         s->dts_sync_point    = INT_MIN;
 328         s->dts_ref_dts_delta = INT_MIN;
 329         s->pts_dts_delta     = INT_MIN;
 330     }
 331
 332     if (s->flags & PARSER_FLAG_ONCE) {
 333         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 334     }
 335     }
 336
 337     *poutbuf = buf;
 338     *poutbuf_size = buf_size;
 339     return next;
 340 }
 341
 342 static int h264_split(AVCodecContext *avctx,
 343                       const uint8_t *buf, int buf_size)
 344 {
 345     int i;
 346     uint32_t state = -1;
 347     int has_sps= 0;
 348
 349     for(i=0; i<=buf_size; i++){
 350         if((state&0xFFFFFF1F) == 0x107)
 351             has_sps=1;
 352 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 353         }*/
 354         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 355             if(has_sps){
 356                 while(i>4 && buf[i-5]==0) i--;
 357                 return i-4;
 358             }
 359         }
 360         if (i<buf_size)
 361             state= (state<<8) | buf[i];
 362     }
 363     return 0;
 364 }
 365
 366 static void close(AVCodecParserContext *s)
 367 {
 368     H264Context *h = s->priv_data;
 369     ParseContext *pc = &h->s.parse_context;
 370
 371     av_free(pc->buffer);
 372     ff_h264_free_context(h);
 373 }
 374
 375 static int init(AVCodecParserContext *s)
 376 {
 377     H264Context *h = s->priv_data;
 378     h->thread_context[0] = h;
 379     h->s.slice_context_count = 1;
 380     return 0;
 381 }
 382
 383 AVCodecParser ff_h264_parser = {
 384     .codec_ids      = { CODEC_ID_H264 },
 385     .priv_data_size = sizeof(H264Context),
 386     .parser_init    = init,
 387     .parser_parse   = h264_parse,
 388     .parser_close   = close,
 389     .split          = h264_split,
 390 };