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 Libav.
   6  *
   7  * Libav 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  * Libav 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 Libav; 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 "libavutil/attributes.h"
  29 #include "parser.h"
  30 #include "h264data.h"
  31 #include "golomb.h"
  32 #include "internal.h"
  33
  34 #include <assert.h>
  35
  36
  37 static int h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_size)
  38 {
  39     int i;
  40     uint32_t state;
  41     ParseContext *pc = &h->parse_context;
  42 //    mb_addr= pc->mb_addr - 1;
  43     state= pc->state;
  44     if(state>13)
  45         state= 7;
  46
  47     for(i=0; i<buf_size; i++){
  48         if(state==7){
  49 #if HAVE_FAST_UNALIGNED
  50         /* we check i<buf_size instead of i+3/7 because its simpler
  51          * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
  52          */
  53 #    if HAVE_FAST_64BIT
  54             while(i<buf_size && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
  55                 i+=8;
  56 #    else
  57             while(i<buf_size && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
  58                 i+=4;
  59 #    endif
  60 #endif
  61             for(; i<buf_size; i++){
  62                 if(!buf[i]){
  63                     state=2;
  64                     break;
  65                 }
  66             }
  67         }else if(state<=2){
  68             if(buf[i]==1)   state^= 5; //2->7, 1->4, 0->5
  69             else if(buf[i]) state = 7;
  70             else            state>>=1; //2->1, 1->0, 0->0
  71         }else if(state<=5){
  72             int v= buf[i] & 0x1F;
  73             if(v==6 || v==7 || v==8 || v==9){
  74                 if(pc->frame_start_found){
  75                     i++;
  76                     goto found;
  77                 }
  78             }else if(v==1 || v==2 || v==5){
  79                 if(pc->frame_start_found){
  80                     state+=8;
  81                     continue;
  82                 }else
  83                     pc->frame_start_found = 1;
  84             }
  85             state= 7;
  86         }else{
  87             if(buf[i] & 0x80)
  88                 goto found;
  89             state= 7;
  90         }
  91     }
  92     pc->state= state;
  93     return END_NOT_FOUND;
  94
  95 found:
  96     pc->state=7;
  97     pc->frame_start_found= 0;
  98     return i-(state&5);
  99 }
 100
 101 /**
 102  * Parse NAL units of found picture and decode some basic information.
 103  *
 104  * @param s parser context.
 105  * @param avctx codec context.
 106  * @param buf buffer with field/frame data.
 107  * @param buf_size size of the buffer.
 108  */
 109 static inline int parse_nal_units(AVCodecParserContext *s,
 110                                   AVCodecContext *avctx,
 111                                   const uint8_t *buf, int buf_size)
 112 {
 113     H264Context *h = s->priv_data;
 114     const uint8_t *buf_end = buf + buf_size;
 115     unsigned int pps_id;
 116     unsigned int slice_type;
 117     int state = -1;
 118     const uint8_t *ptr;
 119
 120     /* set some sane default values */
 121     s->pict_type = AV_PICTURE_TYPE_I;
 122     s->key_frame = 0;
 123
 124     h->avctx= avctx;
 125     h->sei_recovery_frame_cnt = -1;
 126     h->sei_dpb_output_delay         =  0;
 127     h->sei_cpb_removal_delay        = -1;
 128     h->sei_buffering_period_present =  0;
 129
 130     if (!buf_size)
 131         return 0;
 132
 133     for(;;) {
 134         int src_length, dst_length, consumed;
 135         buf = avpriv_find_start_code(buf, buf_end, &state);
 136         if(buf >= buf_end)
 137             break;
 138         --buf;
 139         src_length = buf_end - buf;
 140         switch (state & 0x1f) {
 141         case NAL_SLICE:
 142         case NAL_IDR_SLICE:
 143             // Do not walk the whole buffer just to decode slice header
 144             if (src_length > 20)
 145                 src_length = 20;
 146             break;
 147         }
 148         ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 149         if (ptr==NULL || dst_length < 0)
 150             break;
 151
 152         init_get_bits(&h->gb, ptr, 8*dst_length);
 153         switch(h->nal_unit_type) {
 154         case NAL_SPS:
 155             ff_h264_decode_seq_parameter_set(h);
 156             break;
 157         case NAL_PPS:
 158             ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
 159             break;
 160         case NAL_SEI:
 161             ff_h264_decode_sei(h);
 162             break;
 163         case NAL_IDR_SLICE:
 164             s->key_frame = 1;
 165             /* fall through */
 166         case NAL_SLICE:
 167             get_ue_golomb(&h->gb);  // skip first_mb_in_slice
 168             slice_type = get_ue_golomb_31(&h->gb);
 169             s->pict_type = golomb_to_pict_type[slice_type % 5];
 170             if (h->sei_recovery_frame_cnt >= 0) {
 171                 /* key frame, since recovery_frame_cnt is set */
 172                 s->key_frame = 1;
 173             }
 174             pps_id= get_ue_golomb(&h->gb);
 175             if(pps_id>=MAX_PPS_COUNT) {
 176                 av_log(h->avctx, AV_LOG_ERROR, "pps_id out of range\n");
 177                 return -1;
 178             }
 179             if(!h->pps_buffers[pps_id]) {
 180                 av_log(h->avctx, AV_LOG_ERROR, "non-existing PPS referenced\n");
 181                 return -1;
 182             }
 183             h->pps= *h->pps_buffers[pps_id];
 184             if(!h->sps_buffers[h->pps.sps_id]) {
 185                 av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS referenced\n");
 186                 return -1;
 187             }
 188             h->sps = *h->sps_buffers[h->pps.sps_id];
 189             h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
 190
 191             avctx->profile = ff_h264_get_profile(&h->sps);
 192             avctx->level   = h->sps.level_idc;
 193
 194             if(h->sps.frame_mbs_only_flag){
 195                 h->picture_structure= PICT_FRAME;
 196             }else{
 197                 if(get_bits1(&h->gb)) { //field_pic_flag
 198                     h->picture_structure= PICT_TOP_FIELD + get_bits1(&h->gb); //bottom_field_flag
 199                 } else {
 200                     h->picture_structure= PICT_FRAME;
 201                 }
 202             }
 203
 204             if(h->sps.pic_struct_present_flag) {
 205                 switch (h->sei_pic_struct) {
 206                     case SEI_PIC_STRUCT_TOP_FIELD:
 207                     case SEI_PIC_STRUCT_BOTTOM_FIELD:
 208                         s->repeat_pict = 0;
 209                         break;
 210                     case SEI_PIC_STRUCT_FRAME:
 211                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 212                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 213                         s->repeat_pict = 1;
 214                         break;
 215                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 216                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 217                         s->repeat_pict = 2;
 218                         break;
 219                     case SEI_PIC_STRUCT_FRAME_DOUBLING:
 220                         s->repeat_pict = 3;
 221                         break;
 222                     case SEI_PIC_STRUCT_FRAME_TRIPLING:
 223                         s->repeat_pict = 5;
 224                         break;
 225                     default:
 226                         s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 227                         break;
 228                 }
 229             } else {
 230                 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 231             }
 232
 233             return 0; /* no need to evaluate the rest */
 234         }
 235         buf += consumed;
 236     }
 237     /* didn't find a picture! */
 238     av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n");
 239     return -1;
 240 }
 241
 242 static int h264_parse(AVCodecParserContext *s,
 243                       AVCodecContext *avctx,
 244                       const uint8_t **poutbuf, int *poutbuf_size,
 245                       const uint8_t *buf, int buf_size)
 246 {
 247     H264Context *h = s->priv_data;
 248     ParseContext *pc = &h->parse_context;
 249     int next;
 250
 251     if (!h->got_first) {
 252         h->got_first = 1;
 253         if (avctx->extradata_size) {
 254             h->avctx = avctx;
 255             // must be done like in the decoder.
 256             // otherwise opening the parser, creating extradata,
 257             // and then closing and opening again
 258             // will cause has_b_frames to be always set.
 259             // NB: estimate_timings_from_pts behaves exactly like this.
 260             if (!avctx->has_b_frames)
 261                 h->low_delay = 1;
 262             ff_h264_decode_extradata(h);
 263         }
 264     }
 265
 266     if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){
 267         next= buf_size;
 268     }else{
 269         next = h264_find_frame_end(h, buf, buf_size);
 270
 271         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 272             *poutbuf = NULL;
 273             *poutbuf_size = 0;
 274             return buf_size;
 275         }
 276
 277         if(next<0 && next != END_NOT_FOUND){
 278             assert(pc->last_index + next >= 0 );
 279             h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
 280         }
 281     }
 282
 283     parse_nal_units(s, avctx, buf, buf_size);
 284
 285     if (h->sei_cpb_removal_delay >= 0) {
 286         s->dts_sync_point    = h->sei_buffering_period_present;
 287         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 288         s->pts_dts_delta     = h->sei_dpb_output_delay;
 289     } else {
 290         s->dts_sync_point    = INT_MIN;
 291         s->dts_ref_dts_delta = INT_MIN;
 292         s->pts_dts_delta     = INT_MIN;
 293     }
 294
 295     if (s->flags & PARSER_FLAG_ONCE) {
 296         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 297     }
 298
 299     *poutbuf = buf;
 300     *poutbuf_size = buf_size;
 301     return next;
 302 }
 303
 304 static int h264_split(AVCodecContext *avctx,
 305                       const uint8_t *buf, int buf_size)
 306 {
 307     int i;
 308     uint32_t state = -1;
 309     int has_sps= 0;
 310
 311     for(i=0; i<=buf_size; i++){
 312         if((state&0xFFFFFF1F) == 0x107)
 313             has_sps=1;
 314 /*        if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
 315         }*/
 316         if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
 317             if(has_sps){
 318                 while(i>4 && buf[i-5]==0) i--;
 319                 return i-4;
 320             }
 321         }
 322         if (i<buf_size)
 323             state= (state<<8) | buf[i];
 324     }
 325     return 0;
 326 }
 327
 328 static void close(AVCodecParserContext *s)
 329 {
 330     H264Context *h = s->priv_data;
 331     ParseContext *pc = &h->parse_context;
 332
 333     av_free(pc->buffer);
 334     ff_h264_free_context(h);
 335 }
 336
 337 static av_cold int init(AVCodecParserContext *s)
 338 {
 339     H264Context *h = s->priv_data;
 340     h->thread_context[0] = h;
 341     h->slice_context_count = 1;
 342     return 0;
 343 }
 344
 345 AVCodecParser ff_h264_parser = {
 346     .codec_ids      = { AV_CODEC_ID_H264 },
 347     .priv_data_size = sizeof(H264Context),
 348     .parser_init    = init,
 349     .parser_parse   = h264_parse,
 350     .parser_close   = close,
 351     .split          = h264_split,
 352 };