git.sesse.net Git - ffmpeg/blob - libavcodec/h264_sei.c

   1 /*
   2  * H.26L/H.264/AVC/JVT/14496-10/... sei decoding
   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 sei decoding.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #include "avcodec.h"
  29 #include "golomb.h"
  30 #include "h264.h"
  31 #include "internal.h"
  32
  33 static const uint8_t sei_num_clock_ts_table[9] = {
  34     1, 1, 1, 2, 2, 3, 3, 2, 3
  35 };
  36
  37 void ff_h264_reset_sei(H264Context *h)
  38 {
  39     h->sei_recovery_frame_cnt       = -1;
  40     h->sei_dpb_output_delay         =  0;
  41     h->sei_cpb_removal_delay        = -1;
  42     h->sei_buffering_period_present =  0;
  43     h->sei_frame_packing_present    =  0;
  44 }
  45
  46 static int decode_picture_timing(H264Context *h)
  47 {
  48     SPS *sps = &h->sps;
  49     int i;
  50
  51     for (i = 0; i<MAX_SPS_COUNT; i++)
  52         if (!sps->log2_max_frame_num && h->sps_buffers[i])
  53             sps = h->sps_buffers[i];
  54
  55     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
  56         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
  57                                                  sps->cpb_removal_delay_length);
  58         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
  59                                                  sps->dpb_output_delay_length);
  60     }
  61     if (sps->pic_struct_present_flag) {
  62         unsigned int i, num_clock_ts;
  63
  64         h->sei_pic_struct = get_bits(&h->gb, 4);
  65         h->sei_ct_type    = 0;
  66
  67         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
  68             return AVERROR_INVALIDDATA;
  69
  70         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
  71
  72         for (i = 0; i < num_clock_ts; i++) {
  73             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
  74                 unsigned int full_timestamp_flag;
  75
  76                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
  77                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
  78                 skip_bits(&h->gb, 5);                 /* counting_type */
  79                 full_timestamp_flag = get_bits(&h->gb, 1);
  80                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
  81                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
  82                 skip_bits(&h->gb, 8);                 /* n_frames */
  83                 if (full_timestamp_flag) {
  84                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
  85                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
  86                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
  87                 } else {
  88                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
  89                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
  90                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
  91                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
  92                             if (get_bits(&h->gb, 1))  /* hours_flag */
  93                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
  94                         }
  95                     }
  96                 }
  97                 if (sps->time_offset_length > 0)
  98                     skip_bits(&h->gb,
  99                               sps->time_offset_length); /* time_offset */
 100             }
 101         }
 102
 103         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 104             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
 105                    h->sei_ct_type, h->sei_pic_struct);
 106     }
 107     return 0;
 108 }
 109
 110 static int decode_user_data_itu_t_t35(H264Context *h, int size)
 111 {
 112     uint32_t user_identifier;
 113     int dtg_active_format;
 114
 115     if (size < 7)
 116         return -1;
 117     size -= 7;
 118
 119     skip_bits(&h->gb, 8);   // country_code
 120     skip_bits(&h->gb, 16);  // provider_code
 121     user_identifier = get_bits_long(&h->gb, 32);
 122
 123     switch (user_identifier) {
 124         case 0x44544731:    // "DTG1" - AFD_data
 125             if (size < 1)
 126                 return -1;
 127             skip_bits(&h->gb, 1);
 128             if (get_bits(&h->gb, 1)) {
 129                 skip_bits(&h->gb, 6);
 130                 if (size < 2)
 131                     return -1;
 132                 skip_bits(&h->gb, 4);
 133                 dtg_active_format = get_bits(&h->gb, 4);
 134                 h->avctx->dtg_active_format = dtg_active_format;
 135             } else {
 136                 skip_bits(&h->gb, 6);
 137             }
 138             break;
 139         default:
 140             skip_bits(&h->gb, size * 8);
 141             break;
 142     }
 143
 144     return 0;
 145 }
 146
 147 static int decode_unregistered_user_data(H264Context *h, int size)
 148 {
 149     uint8_t user_data[16 + 256];
 150     int e, build, i;
 151
 152     if (size < 16)
 153         return AVERROR_INVALIDDATA;
 154
 155     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
 156         user_data[i] = get_bits(&h->gb, 8);
 157
 158     user_data[i] = 0;
 159     e = sscanf(user_data + 16, "x264 - core %d", &build);
 160     if (e == 1 && build > 0)
 161         h->x264_build = build;
 162     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
 163         h->x264_build = 67;
 164
 165     if (h->avctx->debug & FF_DEBUG_BUGS)
 166         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
 167
 168     for (; i < size; i++)
 169         skip_bits(&h->gb, 8);
 170
 171     return 0;
 172 }
 173
 174 static int decode_recovery_point(H264Context *h)
 175 {
 176     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
 177
 178     /* 1b exact_match_flag,
 179      * 1b broken_link_flag,
 180      * 2b changing_slice_group_idc */
 181     skip_bits(&h->gb, 4);
 182
 183     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 184         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
 185
 186     return 0;
 187 }
 188
 189 static int decode_buffering_period(H264Context *h)
 190 {
 191     unsigned int sps_id;
 192     int sched_sel_idx;
 193     SPS *sps;
 194
 195     sps_id = get_ue_golomb_31(&h->gb);
 196     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
 197         av_log(h->avctx, AV_LOG_ERROR,
 198                "non-existing SPS %d referenced in buffering period\n", sps_id);
 199         return AVERROR_INVALIDDATA;
 200     }
 201     sps = h->sps_buffers[sps_id];
 202
 203     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
 204     if (sps->nal_hrd_parameters_present_flag) {
 205         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 206             h->initial_cpb_removal_delay[sched_sel_idx] =
 207                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 208             // initial_cpb_removal_delay_offset
 209             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 210         }
 211     }
 212     if (sps->vcl_hrd_parameters_present_flag) {
 213         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 214             h->initial_cpb_removal_delay[sched_sel_idx] =
 215                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 216             // initial_cpb_removal_delay_offset
 217             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 218         }
 219     }
 220
 221     h->sei_buffering_period_present = 1;
 222     return 0;
 223 }
 224
 225 static int decode_frame_packing_arrangement(H264Context *h)
 226 {
 227     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
 228     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
 229     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
 230
 231     if (h->sei_frame_packing_present) {
 232         h->sei_fpa.frame_packing_arrangement_type =
 233         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
 234         h->sei_fpa.quincunx_sampling_flag         =
 235         h->quincunx_subsampling           = get_bits1(&h->gb);
 236         h->sei_fpa.content_interpretation_type    =
 237         h->content_interpretation_type    = get_bits(&h->gb, 6);
 238
 239         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
 240         // field_views_flag, current_frame_is_frame0_flag,
 241         // frame0_self_contained_flag, frame1_self_contained_flag
 242         skip_bits(&h->gb, 6);
 243
 244         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
 245             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
 246         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
 247         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
 248     }
 249     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
 250
 251     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 252         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
 253                                        h->sei_fpa.frame_packing_arrangement_id,
 254                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
 255                                        h->sei_fpa.frame_packing_arrangement_type,
 256                                        h->sei_fpa.quincunx_sampling_flag,
 257                                        h->sei_fpa.content_interpretation_type,
 258                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
 259
 260     return 0;
 261 }
 262
 263 int ff_h264_decode_sei(H264Context *h)
 264 {
 265     while (get_bits_left(&h->gb) > 16) {
 266         int type = 0;
 267         unsigned size = 0;
 268         unsigned next;
 269         int ret  = 0;
 270
 271         do {
 272             if (get_bits_left(&h->gb) < 8)
 273                 return AVERROR_INVALIDDATA;
 274             type += show_bits(&h->gb, 8);
 275         } while (get_bits(&h->gb, 8) == 255);
 276
 277         do {
 278             if (get_bits_left(&h->gb) < 8)
 279                 return AVERROR_INVALIDDATA;
 280             size += show_bits(&h->gb, 8);
 281         } while (get_bits(&h->gb, 8) == 255);
 282
 283         if (h->avctx->debug&FF_DEBUG_STARTCODE)
 284             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
 285
 286         if (size > get_bits_left(&h->gb) / 8) {
 287             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
 288                    type, 8*size, get_bits_left(&h->gb));
 289             return AVERROR_INVALIDDATA;
 290         }
 291         next = get_bits_count(&h->gb) + 8 * size;
 292
 293         switch (type) {
 294         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
 295             ret = decode_picture_timing(h);
 296             if (ret < 0)
 297                 return ret;
 298             break;
 299         case SEI_TYPE_USER_DATA_ITU_T_T35:
 300             if (decode_user_data_itu_t_t35(h, size) < 0)
 301                 return -1;
 302             break;
 303         case SEI_TYPE_USER_DATA_UNREGISTERED:
 304             ret = decode_unregistered_user_data(h, size);
 305             if (ret < 0)
 306                 return ret;
 307             break;
 308         case SEI_TYPE_RECOVERY_POINT:
 309             ret = decode_recovery_point(h);
 310             if (ret < 0)
 311                 return ret;
 312             break;
 313         case SEI_TYPE_BUFFERING_PERIOD:
 314             ret = decode_buffering_period(h);
 315             if (ret < 0)
 316                 return ret;
 317             break;
 318         case SEI_TYPE_FRAME_PACKING:
 319             ret = decode_frame_packing_arrangement(h);
 320             if (ret < 0)
 321                 return ret;
 322             break;
 323         default:
 324             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
 325         }
 326         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
 327
 328         // FIXME check bits here
 329         align_get_bits(&h->gb);
 330     }
 331
 332     return 0;
 333 }
 334
 335 const char* ff_h264_sei_stereo_mode(H264Context *h)
 336 {
 337     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
 338         switch (h->sei_fpa.frame_packing_arrangement_type) {
 339             case SEI_FPA_TYPE_CHECKERBOARD:
 340                 if (h->sei_fpa.content_interpretation_type == 2)
 341                     return "checkerboard_rl";
 342                 else
 343                     return "checkerboard_lr";
 344             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
 345                 if (h->sei_fpa.content_interpretation_type == 2)
 346                     return "col_interleaved_rl";
 347                 else
 348                     return "col_interleaved_lr";
 349             case SEI_FPA_TYPE_INTERLEAVE_ROW:
 350                 if (h->sei_fpa.content_interpretation_type == 2)
 351                     return "row_interleaved_rl";
 352                 else
 353                     return "row_interleaved_lr";
 354             case SEI_FPA_TYPE_SIDE_BY_SIDE:
 355                 if (h->sei_fpa.content_interpretation_type == 2)
 356                     return "right_left";
 357                 else
 358                     return "left_right";
 359             case SEI_FPA_TYPE_TOP_BOTTOM:
 360                 if (h->sei_fpa.content_interpretation_type == 2)
 361                     return "bottom_top";
 362                 else
 363                     return "top_bottom";
 364             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
 365                 if (h->sei_fpa.content_interpretation_type == 2)
 366                     return "block_rl";
 367                 else
 368                     return "block_lr";
 369             case SEI_FPA_TYPE_2D:
 370             default:
 371                 return "mono";
 372         }
 373     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
 374         return "mono";
 375     } else {
 376         return NULL;
 377     }
 378 }