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 "internal.h"
  29 #include "avcodec.h"
  30 #include "h264.h"
  31 #include "golomb.h"
  32
  33 #include <assert.h>
  34
  35 static const uint8_t sei_num_clock_ts_table[9] = {
  36     1, 1, 1, 2, 2, 3, 3, 2, 3
  37 };
  38
  39 void ff_h264_reset_sei(H264Context *h)
  40 {
  41     h->sei_recovery_frame_cnt       = -1;
  42     h->sei_dpb_output_delay         =  0;
  43     h->sei_cpb_removal_delay        = -1;
  44     h->sei_buffering_period_present =  0;
  45     h->sei_frame_packing_present    =  0;
  46 }
  47
  48 static int decode_picture_timing(H264Context *h)
  49 {
  50     SPS *sps = &h->sps;
  51     int i;
  52
  53     for (i = 0; i<MAX_SPS_COUNT; i++)
  54         if (!sps->log2_max_frame_num && h->sps_buffers[i])
  55             sps = h->sps_buffers[i];
  56
  57     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
  58         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
  59                                                  sps->cpb_removal_delay_length);
  60         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
  61                                                  sps->dpb_output_delay_length);
  62     }
  63     if (sps->pic_struct_present_flag) {
  64         unsigned int i, num_clock_ts;
  65
  66         h->sei_pic_struct = get_bits(&h->gb, 4);
  67         h->sei_ct_type    = 0;
  68
  69         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
  70             return AVERROR_INVALIDDATA;
  71
  72         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
  73
  74         for (i = 0; i < num_clock_ts; i++) {
  75             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
  76                 unsigned int full_timestamp_flag;
  77
  78                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
  79                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
  80                 skip_bits(&h->gb, 5);                 /* counting_type */
  81                 full_timestamp_flag = get_bits(&h->gb, 1);
  82                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
  83                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
  84                 skip_bits(&h->gb, 8);                 /* n_frames */
  85                 if (full_timestamp_flag) {
  86                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
  87                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
  88                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
  89                 } else {
  90                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
  91                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
  92                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
  93                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
  94                             if (get_bits(&h->gb, 1))  /* hours_flag */
  95                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
  96                         }
  97                     }
  98                 }
  99                 if (sps->time_offset_length > 0)
 100                     skip_bits(&h->gb,
 101                               sps->time_offset_length); /* time_offset */
 102             }
 103         }
 104
 105         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 106             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
 107                    h->sei_ct_type, h->sei_pic_struct);
 108     }
 109     return 0;
 110 }
 111
 112 static int decode_user_data_itu_t_t35(H264Context *h, int size)
 113 {
 114     uint32_t user_identifier;
 115     int dtg_active_format;
 116
 117     if (size < 7)
 118         return -1;
 119     size -= 7;
 120
 121     skip_bits(&h->gb, 8);   // country_code
 122     skip_bits(&h->gb, 16);  // provider_code
 123     user_identifier = get_bits_long(&h->gb, 32);
 124
 125     switch (user_identifier) {
 126         case 0x44544731:    // "DTG1" - AFD_data
 127             if (size < 1)
 128                 return -1;
 129             skip_bits(&h->gb, 1);
 130             if (get_bits(&h->gb, 1)) {
 131                 skip_bits(&h->gb, 6);
 132                 if (size < 2)
 133                     return -1;
 134                 skip_bits(&h->gb, 4);
 135                 dtg_active_format = get_bits(&h->gb, 4);
 136                 h->avctx->dtg_active_format = dtg_active_format;
 137             } else {
 138                 skip_bits(&h->gb, 6);
 139             }
 140             break;
 141         default:
 142             skip_bits(&h->gb, size * 8);
 143             break;
 144     }
 145
 146     return 0;
 147 }
 148
 149 static int decode_unregistered_user_data(H264Context *h, int size)
 150 {
 151     uint8_t user_data[16 + 256];
 152     int e, build, i;
 153
 154     if (size < 16)
 155         return AVERROR_INVALIDDATA;
 156
 157     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
 158         user_data[i] = get_bits(&h->gb, 8);
 159
 160     user_data[i] = 0;
 161     e = sscanf(user_data + 16, "x264 - core %d", &build);
 162     if (e == 1 && build > 0)
 163         h->x264_build = build;
 164     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
 165         h->x264_build = 67;
 166
 167     if (h->avctx->debug & FF_DEBUG_BUGS)
 168         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
 169
 170     for (; i < size; i++)
 171         skip_bits(&h->gb, 8);
 172
 173     return 0;
 174 }
 175
 176 static int decode_recovery_point(H264Context *h)
 177 {
 178     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
 179
 180     /* 1b exact_match_flag,
 181      * 1b broken_link_flag,
 182      * 2b changing_slice_group_idc */
 183     skip_bits(&h->gb, 4);
 184
 185     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 186         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
 187
 188     return 0;
 189 }
 190
 191 static int decode_buffering_period(H264Context *h)
 192 {
 193     unsigned int sps_id;
 194     int sched_sel_idx;
 195     SPS *sps;
 196
 197     sps_id = get_ue_golomb_31(&h->gb);
 198     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
 199         av_log(h->avctx, AV_LOG_ERROR,
 200                "non-existing SPS %d referenced in buffering period\n", sps_id);
 201         return AVERROR_INVALIDDATA;
 202     }
 203     sps = h->sps_buffers[sps_id];
 204
 205     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
 206     if (sps->nal_hrd_parameters_present_flag) {
 207         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 208             h->initial_cpb_removal_delay[sched_sel_idx] =
 209                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 210             // initial_cpb_removal_delay_offset
 211             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 212         }
 213     }
 214     if (sps->vcl_hrd_parameters_present_flag) {
 215         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 216             h->initial_cpb_removal_delay[sched_sel_idx] =
 217                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 218             // initial_cpb_removal_delay_offset
 219             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 220         }
 221     }
 222
 223     h->sei_buffering_period_present = 1;
 224     return 0;
 225 }
 226
 227 static int decode_frame_packing_arrangement(H264Context *h)
 228 {
 229     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
 230     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
 231     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
 232
 233     if (h->sei_frame_packing_present) {
 234         h->sei_fpa.frame_packing_arrangement_type =
 235         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
 236         h->sei_fpa.quincunx_sampling_flag         =
 237         h->quincunx_subsampling           = get_bits1(&h->gb);
 238         h->sei_fpa.content_interpretation_type    =
 239         h->content_interpretation_type    = get_bits(&h->gb, 6);
 240
 241         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
 242         // field_views_flag, current_frame_is_frame0_flag,
 243         // frame0_self_contained_flag, frame1_self_contained_flag
 244         skip_bits(&h->gb, 6);
 245
 246         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
 247             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
 248         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
 249         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
 250     }
 251     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
 252
 253     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 254         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
 255                                        h->sei_fpa.frame_packing_arrangement_id,
 256                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
 257                                        h->sei_fpa.frame_packing_arrangement_type,
 258                                        h->sei_fpa.quincunx_sampling_flag,
 259                                        h->sei_fpa.content_interpretation_type,
 260                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
 261
 262     return 0;
 263 }
 264
 265 int ff_h264_decode_sei(H264Context *h)
 266 {
 267     while (get_bits_left(&h->gb) > 16) {
 268         int type = 0;
 269         unsigned size = 0;
 270         unsigned next;
 271         int ret  = 0;
 272
 273         do {
 274             if (get_bits_left(&h->gb) < 8)
 275                 return AVERROR_INVALIDDATA;
 276             type += show_bits(&h->gb, 8);
 277         } while (get_bits(&h->gb, 8) == 255);
 278
 279         do {
 280             if (get_bits_left(&h->gb) < 8)
 281                 return AVERROR_INVALIDDATA;
 282             size += show_bits(&h->gb, 8);
 283         } while (get_bits(&h->gb, 8) == 255);
 284
 285         if (h->avctx->debug&FF_DEBUG_STARTCODE)
 286             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
 287
 288         if (size > get_bits_left(&h->gb) / 8) {
 289             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
 290                    type, 8*size, get_bits_left(&h->gb));
 291             return AVERROR_INVALIDDATA;
 292         }
 293         next = get_bits_count(&h->gb) + 8 * size;
 294
 295         switch (type) {
 296         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
 297             ret = decode_picture_timing(h);
 298             if (ret < 0)
 299                 return ret;
 300             break;
 301         case SEI_TYPE_USER_DATA_ITU_T_T35:
 302             if (decode_user_data_itu_t_t35(h, size) < 0)
 303                 return -1;
 304             break;
 305         case SEI_TYPE_USER_DATA_UNREGISTERED:
 306             ret = decode_unregistered_user_data(h, size);
 307             if (ret < 0)
 308                 return ret;
 309             break;
 310         case SEI_TYPE_RECOVERY_POINT:
 311             ret = decode_recovery_point(h);
 312             if (ret < 0)
 313                 return ret;
 314             break;
 315         case SEI_BUFFERING_PERIOD:
 316             ret = decode_buffering_period(h);
 317             if (ret < 0)
 318                 return ret;
 319             break;
 320         case SEI_TYPE_FRAME_PACKING:
 321             ret = decode_frame_packing_arrangement(h);
 322             if (ret < 0)
 323                 return ret;
 324             break;
 325         default:
 326             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
 327         }
 328         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
 329
 330         // FIXME check bits here
 331         align_get_bits(&h->gb);
 332     }
 333
 334     return 0;
 335 }
 336
 337 const char* ff_h264_sei_stereo_mode(H264Context *h)
 338 {
 339     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
 340         switch (h->sei_fpa.frame_packing_arrangement_type) {
 341             case SEI_FPA_TYPE_CHECKERBOARD:
 342                 if (h->sei_fpa.content_interpretation_type == 2)
 343                     return "checkerboard_rl";
 344                 else
 345                     return "checkerboard_lr";
 346             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
 347                 if (h->sei_fpa.content_interpretation_type == 2)
 348                     return "col_interleaved_rl";
 349                 else
 350                     return "col_interleaved_lr";
 351             case SEI_FPA_TYPE_INTERLEAVE_ROW:
 352                 if (h->sei_fpa.content_interpretation_type == 2)
 353                     return "row_interleaved_rl";
 354                 else
 355                     return "row_interleaved_lr";
 356             case SEI_FPA_TYPE_SIDE_BY_SIDE:
 357                 if (h->sei_fpa.content_interpretation_type == 2)
 358                     return "right_left";
 359                 else
 360                     return "left_right";
 361             case SEI_FPA_TYPE_TOP_BOTTOM:
 362                 if (h->sei_fpa.content_interpretation_type == 2)
 363                     return "bottom_top";
 364                 else
 365                     return "top_bottom";
 366             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
 367                 if (h->sei_fpa.content_interpretation_type == 2)
 368                     return "block_rl";
 369                 else
 370                     return "block_lr";
 371             case SEI_FPA_TYPE_2D:
 372             default:
 373                 return "mono";
 374         }
 375     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
 376         return "mono";
 377     } else {
 378         return NULL;
 379     }
 380 }