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,
  38                                int buf_size)
  39 {
  40     int i;
  41     uint32_t state;
  42     ParseContext *pc = &h->parse_context;
  43 //    mb_addr= pc->mb_addr - 1;
  44     state = pc->state;
  45     if (state > 13)
  46         state = 7;
  47
  48     for (i = 0; i < buf_size; i++) {
  49         if (state == 7) {
  50 #if HAVE_FAST_UNALIGNED
  51             /* we check i < buf_size instead of i + 3 / 7 because it is
  52              * simpler and there must be FF_INPUT_BUFFER_PADDING_SIZE
  53              * bytes at the end.
  54              */
  55 #if HAVE_FAST_64BIT
  56             while (i < buf_size &&
  57                    !((~*(const uint64_t *)(buf + i) &
  58                       (*(const uint64_t *)(buf + i) - 0x0101010101010101ULL)) &
  59                       0x8080808080808080ULL))
  60                 i += 8;
  61 #else
  62             while (i < buf_size &&
  63                    !((~*(const uint32_t *)(buf + i) &
  64                       (*(const uint32_t *)(buf + i) - 0x01010101U)) &
  65                       0x80808080U))
  66                 i += 4;
  67 #endif
  68 #endif
  69             for (; i < buf_size; i++)
  70                 if (!buf[i]) {
  71                     state = 2;
  72                     break;
  73                 }
  74         } else if (state <= 2) {
  75             if (buf[i] == 1)
  76                 state ^= 5;            // 2->7, 1->4, 0->5
  77             else if (buf[i])
  78                 state = 7;
  79             else
  80                 state >>= 1;           // 2->1, 1->0, 0->0
  81         } else if (state <= 5) {
  82             int v = buf[i] & 0x1F;
  83             if (v == 6 || v == 7 || v == 8 || v == 9) {
  84                 if (pc->frame_start_found) {
  85                     i++;
  86                     goto found;
  87                 }
  88             } else if (v == 1 || v == 2 || v == 5) {
  89                 if (pc->frame_start_found) {
  90                     state += 8;
  91                     continue;
  92                 } else
  93                     pc->frame_start_found = 1;
  94             }
  95             state = 7;
  96         } else {
  97             if (buf[i] & 0x80)
  98                 goto found;
  99             state = 7;
 100         }
 101     }
 102     pc->state = state;
 103     return END_NOT_FOUND;
 104
 105 found:
 106     pc->state             = 7;
 107     pc->frame_start_found = 0;
 108     return i - (state & 5);
 109 }
 110
 111 /**
 112  * Parse NAL units of found picture and decode some basic information.
 113  *
 114  * @param s parser context.
 115  * @param avctx codec context.
 116  * @param buf buffer with field/frame data.
 117  * @param buf_size size of the buffer.
 118  */
 119 static inline int parse_nal_units(AVCodecParserContext *s,
 120                                   AVCodecContext *avctx,
 121                                   const uint8_t *buf, int buf_size)
 122 {
 123     H264Context *h         = s->priv_data;
 124     const uint8_t *buf_end = buf + buf_size;
 125     unsigned int pps_id;
 126     unsigned int slice_type;
 127     int state = -1;
 128     const uint8_t *ptr;
 129     int field_poc[2];
 130
 131     /* set some sane default values */
 132     s->pict_type         = AV_PICTURE_TYPE_I;
 133     s->key_frame         = 0;
 134     s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
 135
 136     h->avctx                        = avctx;
 137     h->sei_recovery_frame_cnt       = -1;
 138     h->sei_dpb_output_delay         = 0;
 139     h->sei_cpb_removal_delay        = -1;
 140     h->sei_buffering_period_present = 0;
 141
 142     if (!buf_size)
 143         return 0;
 144
 145     for (;;) {
 146         int src_length, dst_length, consumed;
 147         buf = avpriv_find_start_code(buf, buf_end, &state);
 148         if (buf >= buf_end)
 149             break;
 150         --buf;
 151         src_length = buf_end - buf;
 152         switch (state & 0x1f) {
 153         case NAL_SLICE:
 154         case NAL_IDR_SLICE:
 155             // Do not walk the whole buffer just to decode slice header
 156             if (src_length > 20)
 157                 src_length = 20;
 158             break;
 159         }
 160         ptr = ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length);
 161         if (ptr == NULL || dst_length < 0)
 162             break;
 163
 164         init_get_bits(&h->gb, ptr, 8 * dst_length);
 165         switch (h->nal_unit_type) {
 166         case NAL_SPS:
 167             ff_h264_decode_seq_parameter_set(h);
 168             break;
 169         case NAL_PPS:
 170             ff_h264_decode_picture_parameter_set(h, h->gb.size_in_bits);
 171             break;
 172         case NAL_SEI:
 173             ff_h264_decode_sei(h);
 174             break;
 175         case NAL_IDR_SLICE:
 176             s->key_frame = 1;
 177
 178             h->prev_frame_num        = 0;
 179             h->prev_frame_num_offset = 0;
 180             h->prev_poc_msb          =
 181             h->prev_poc_lsb          = 0;
 182         /* fall through */
 183         case NAL_SLICE:
 184             get_ue_golomb(&h->gb);  // skip first_mb_in_slice
 185             slice_type   = get_ue_golomb_31(&h->gb);
 186             s->pict_type = golomb_to_pict_type[slice_type % 5];
 187             if (h->sei_recovery_frame_cnt >= 0) {
 188                 /* key frame, since recovery_frame_cnt is set */
 189                 s->key_frame = 1;
 190             }
 191             pps_id = get_ue_golomb(&h->gb);
 192             if (pps_id >= MAX_PPS_COUNT) {
 193                 av_log(h->avctx, AV_LOG_ERROR,
 194                        "pps_id out of range\n");
 195                 return -1;
 196             }
 197             if (!h->pps_buffers[pps_id]) {
 198                 av_log(h->avctx, AV_LOG_ERROR,
 199                        "non-existing PPS referenced\n");
 200                 return -1;
 201             }
 202             h->pps = *h->pps_buffers[pps_id];
 203             if (!h->sps_buffers[h->pps.sps_id]) {
 204                 av_log(h->avctx, AV_LOG_ERROR,
 205                        "non-existing SPS referenced\n");
 206                 return -1;
 207             }
 208             h->sps       = *h->sps_buffers[h->pps.sps_id];
 209             h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num);
 210
 211             avctx->profile = ff_h264_get_profile(&h->sps);
 212             avctx->level   = h->sps.level_idc;
 213
 214             if (h->sps.frame_mbs_only_flag) {
 215                 h->picture_structure = PICT_FRAME;
 216             } else {
 217                 if (get_bits1(&h->gb)) { // field_pic_flag
 218                     h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
 219                 } else {
 220                     h->picture_structure = PICT_FRAME;
 221                 }
 222             }
 223
 224             if (h->nal_unit_type == NAL_IDR_SLICE)
 225                 get_ue_golomb(&h->gb); /* idr_pic_id */
 226             if (h->sps.poc_type == 0) {
 227                 h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
 228
 229                 if (h->pps.pic_order_present == 1 &&
 230                     h->picture_structure == PICT_FRAME)
 231                     h->delta_poc_bottom = get_se_golomb(&h->gb);
 232             }
 233
 234             if (h->sps.poc_type == 1 &&
 235                 !h->sps.delta_pic_order_always_zero_flag) {
 236                 h->delta_poc[0] = get_se_golomb(&h->gb);
 237
 238                 if (h->pps.pic_order_present == 1 &&
 239                     h->picture_structure == PICT_FRAME)
 240                     h->delta_poc[1] = get_se_golomb(&h->gb);
 241             }
 242
 243             ff_init_poc(h, field_poc, &s->output_picture_number);
 244
 245             if (h->sps.pic_struct_present_flag) {
 246                 switch (h->sei_pic_struct) {
 247                 case SEI_PIC_STRUCT_TOP_FIELD:
 248                 case SEI_PIC_STRUCT_BOTTOM_FIELD:
 249                     s->repeat_pict = 0;
 250                     break;
 251                 case SEI_PIC_STRUCT_FRAME:
 252                 case SEI_PIC_STRUCT_TOP_BOTTOM:
 253                 case SEI_PIC_STRUCT_BOTTOM_TOP:
 254                     s->repeat_pict = 1;
 255                     break;
 256                 case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 257                 case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 258                     s->repeat_pict = 2;
 259                     break;
 260                 case SEI_PIC_STRUCT_FRAME_DOUBLING:
 261                     s->repeat_pict = 3;
 262                     break;
 263                 case SEI_PIC_STRUCT_FRAME_TRIPLING:
 264                     s->repeat_pict = 5;
 265                     break;
 266                 default:
 267                     s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 268                     break;
 269                 }
 270             } else {
 271                 s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
 272             }
 273
 274             if (h->picture_structure == PICT_FRAME) {
 275                 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
 276                 if (h->sps.pic_struct_present_flag) {
 277                     switch (h->sei_pic_struct) {
 278                     case SEI_PIC_STRUCT_TOP_BOTTOM:
 279                     case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 280                         s->field_order = AV_FIELD_TT;
 281                         break;
 282                     case SEI_PIC_STRUCT_BOTTOM_TOP:
 283                     case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 284                         s->field_order = AV_FIELD_BB;
 285                         break;
 286                     default:
 287                         s->field_order = AV_FIELD_PROGRESSIVE;
 288                         break;
 289                     }
 290                 } else {
 291                     if (field_poc[0] < field_poc[1])
 292                         s->field_order = AV_FIELD_TT;
 293                     else if (field_poc[0] > field_poc[1])
 294                         s->field_order = AV_FIELD_BB;
 295                     else
 296                         s->field_order = AV_FIELD_PROGRESSIVE;
 297                 }
 298             } else {
 299                 if (h->picture_structure == PICT_TOP_FIELD)
 300                     s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
 301                 else
 302                     s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
 303                 s->field_order = AV_FIELD_UNKNOWN;
 304             }
 305
 306             return 0; /* no need to evaluate the rest */
 307         }
 308         buf += consumed;
 309     }
 310     /* didn't find a picture! */
 311     av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n");
 312     return -1;
 313 }
 314
 315 static int h264_parse(AVCodecParserContext *s,
 316                       AVCodecContext *avctx,
 317                       const uint8_t **poutbuf, int *poutbuf_size,
 318                       const uint8_t *buf, int buf_size)
 319 {
 320     H264Context *h   = s->priv_data;
 321     ParseContext *pc = &h->parse_context;
 322     int next;
 323
 324     if (!h->got_first) {
 325         h->got_first = 1;
 326         if (avctx->extradata_size) {
 327             h->avctx = avctx;
 328             // must be done like in the decoder.
 329             // otherwise opening the parser, creating extradata,
 330             // and then closing and opening again
 331             // will cause has_b_frames to be always set.
 332             // NB: estimate_timings_from_pts behaves exactly like this.
 333             if (!avctx->has_b_frames)
 334                 h->low_delay = 1;
 335             ff_h264_decode_extradata(h);
 336         }
 337     }
 338
 339     if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
 340         next = buf_size;
 341     } else {
 342         next = h264_find_frame_end(h, buf, buf_size);
 343
 344         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 345             *poutbuf      = NULL;
 346             *poutbuf_size = 0;
 347             return buf_size;
 348         }
 349
 350         if (next < 0 && next != END_NOT_FOUND) {
 351             assert(pc->last_index + next >= 0);
 352             h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
 353         }
 354     }
 355
 356     parse_nal_units(s, avctx, buf, buf_size);
 357
 358     if (h->sei_cpb_removal_delay >= 0) {
 359         s->dts_sync_point    = h->sei_buffering_period_present;
 360         s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
 361         s->pts_dts_delta     = h->sei_dpb_output_delay;
 362     } else {
 363         s->dts_sync_point    = INT_MIN;
 364         s->dts_ref_dts_delta = INT_MIN;
 365         s->pts_dts_delta     = INT_MIN;
 366     }
 367
 368     if (s->flags & PARSER_FLAG_ONCE) {
 369         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 370     }
 371
 372     *poutbuf      = buf;
 373     *poutbuf_size = buf_size;
 374     return next;
 375 }
 376
 377 static int h264_split(AVCodecContext *avctx,
 378                       const uint8_t *buf, int buf_size)
 379 {
 380     int i;
 381     uint32_t state = -1;
 382     int has_sps    = 0;
 383
 384     for (i = 0; i <= buf_size; i++) {
 385         if ((state & 0xFFFFFF1F) == 0x107)
 386             has_sps = 1;
 387         /*  if((state&0xFFFFFF1F) == 0x101 ||
 388          *     (state&0xFFFFFF1F) == 0x102 ||
 389          *     (state&0xFFFFFF1F) == 0x105) {
 390          *  }
 391          */
 392         if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x107 &&
 393             (state & 0xFFFFFF1F) != 0x108 && (state & 0xFFFFFF1F) != 0x109) {
 394             if (has_sps) {
 395                 while (i > 4 && buf[i - 5] == 0)
 396                     i--;
 397                 return i - 4;
 398             }
 399         }
 400         if (i < buf_size)
 401             state = (state << 8) | buf[i];
 402     }
 403     return 0;
 404 }
 405
 406 static void close(AVCodecParserContext *s)
 407 {
 408     H264Context *h   = s->priv_data;
 409     ParseContext *pc = &h->parse_context;
 410
 411     av_free(pc->buffer);
 412     ff_h264_free_context(h);
 413 }
 414
 415 static av_cold int init(AVCodecParserContext *s)
 416 {
 417     H264Context *h = s->priv_data;
 418     h->thread_context[0]   = h;
 419     h->slice_context_count = 1;
 420     return 0;
 421 }
 422
 423 AVCodecParser ff_h264_parser = {
 424     .codec_ids      = { AV_CODEC_ID_H264 },
 425     .priv_data_size = sizeof(H264Context),
 426     .parser_init    = init,
 427     .parser_parse   = h264_parse,
 428     .parser_close   = close,
 429     .split          = h264_split,
 430 };