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     h->sei_display_orientation_present = 0;
  45     h->sei_reguserdata_afd_present  =  0;
  46
  47     h->a53_caption_size = 0;
  48     av_freep(&h->a53_caption);
  49 }
  50
  51 static int decode_picture_timing(H264Context *h)
  52 {
  53     SPS *sps = &h->sps;
  54     int i;
  55
  56     for (i = 0; i<MAX_SPS_COUNT; i++)
  57         if (!sps->log2_max_frame_num && h->sps_buffers[i])
  58             sps = h->sps_buffers[i];
  59
  60     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
  61         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
  62                                                  sps->cpb_removal_delay_length);
  63         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
  64                                                  sps->dpb_output_delay_length);
  65     }
  66     if (sps->pic_struct_present_flag) {
  67         unsigned int i, num_clock_ts;
  68
  69         h->sei_pic_struct = get_bits(&h->gb, 4);
  70         h->sei_ct_type    = 0;
  71
  72         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
  73             return AVERROR_INVALIDDATA;
  74
  75         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
  76
  77         for (i = 0; i < num_clock_ts; i++) {
  78             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
  79                 unsigned int full_timestamp_flag;
  80
  81                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
  82                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
  83                 skip_bits(&h->gb, 5);                 /* counting_type */
  84                 full_timestamp_flag = get_bits(&h->gb, 1);
  85                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
  86                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
  87                 skip_bits(&h->gb, 8);                 /* n_frames */
  88                 if (full_timestamp_flag) {
  89                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
  90                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
  91                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
  92                 } else {
  93                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
  94                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
  95                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
  96                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
  97                             if (get_bits(&h->gb, 1))  /* hours_flag */
  98                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
  99                         }
 100                     }
 101                 }
 102                 if (sps->time_offset_length > 0)
 103                     skip_bits(&h->gb,
 104                               sps->time_offset_length); /* time_offset */
 105             }
 106         }
 107
 108         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 109             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
 110                    h->sei_ct_type, h->sei_pic_struct);
 111     }
 112     return 0;
 113 }
 114
 115 static int decode_registered_user_data_afd(H264Context *h, int size)
 116 {
 117     int flag;
 118
 119     if (size-- < 1)
 120         return AVERROR_INVALIDDATA;
 121     skip_bits(&h->gb, 1);               // 0
 122     flag = get_bits(&h->gb, 1);         // active_format_flag
 123     skip_bits(&h->gb, 6);               // reserved
 124
 125     if (flag) {
 126         if (size-- < 1)
 127             return AVERROR_INVALIDDATA;
 128         skip_bits(&h->gb, 4);           // reserved
 129         h->active_format_description   = get_bits(&h->gb, 4);
 130         h->sei_reguserdata_afd_present = 1;
 131 #if FF_API_AFD
 132 FF_DISABLE_DEPRECATION_WARNINGS
 133         h->avctx->dtg_active_format = h->active_format_description;
 134 FF_ENABLE_DEPRECATION_WARNINGS
 135 #endif /* FF_API_AFD */
 136     }
 137
 138     return 0;
 139 }
 140
 141 static int decode_registered_user_data_closed_caption(H264Context *h, int size)
 142 {
 143     int flag;
 144     int user_data_type_code;
 145     int cc_count;
 146
 147     if (size < 3)
 148         return AVERROR(EINVAL);
 149
 150     user_data_type_code = get_bits(&h->gb, 8);
 151     if (user_data_type_code == 0x3) {
 152         skip_bits(&h->gb, 1);           // reserved
 153
 154         flag = get_bits(&h->gb, 1);     // process_cc_data_flag
 155         if (flag) {
 156             skip_bits(&h->gb, 1);       // zero bit
 157             cc_count = get_bits(&h->gb, 5);
 158             skip_bits(&h->gb, 8);       // reserved
 159             size -= 2;
 160
 161             if (cc_count && size >= cc_count * 3) {
 162                 const uint64_t new_size = (h->a53_caption_size + cc_count
 163                                            * UINT64_C(3));
 164                 int i, ret;
 165
 166                 if (new_size > INT_MAX)
 167                     return AVERROR(EINVAL);
 168
 169                 /* Allow merging of the cc data from two fields. */
 170                 ret = av_reallocp(&h->a53_caption, new_size);
 171                 if (ret < 0)
 172                     return ret;
 173
 174                 for (i = 0; i < cc_count; i++) {
 175                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 176                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 177                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 178                 }
 179
 180                 skip_bits(&h->gb, 8);   // marker_bits
 181             }
 182         }
 183     }
 184
 185     return 0;
 186 }
 187
 188 static int decode_registered_user_data(H264Context *h, int size)
 189 {
 190     uint32_t country_code;
 191     uint32_t user_identifier;
 192
 193     if (size < 7)
 194         return AVERROR_INVALIDDATA;
 195     size -= 7;
 196
 197     country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
 198     if (country_code == 0xFF) {
 199         skip_bits(&h->gb, 8);           // itu_t_t35_country_code_extension_byte
 200         size--;
 201     }
 202
 203     /* itu_t_t35_payload_byte follows */
 204     skip_bits(&h->gb, 8);              // terminal provider code
 205     skip_bits(&h->gb, 8);              // terminal provider oriented code
 206     user_identifier = get_bits_long(&h->gb, 32);
 207
 208     switch (user_identifier) {
 209         case MKBETAG('D', 'T', 'G', '1'):       // afd_data
 210             return decode_registered_user_data_afd(h, size);
 211         case MKBETAG('G', 'A', '9', '4'):       // closed captions
 212             return decode_registered_user_data_closed_caption(h, size);
 213         default:
 214             skip_bits(&h->gb, size * 8);
 215             break;
 216     }
 217
 218     return 0;
 219 }
 220
 221 static int decode_unregistered_user_data(H264Context *h, int size)
 222 {
 223     uint8_t user_data[16 + 256];
 224     int e, build, i;
 225
 226     if (size < 16)
 227         return AVERROR_INVALIDDATA;
 228
 229     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
 230         user_data[i] = get_bits(&h->gb, 8);
 231
 232     user_data[i] = 0;
 233     e = sscanf(user_data + 16, "x264 - core %d", &build);
 234     if (e == 1 && build > 0)
 235         h->x264_build = build;
 236     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
 237         h->x264_build = 67;
 238
 239     if (h->avctx->debug & FF_DEBUG_BUGS)
 240         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
 241
 242     for (; i < size; i++)
 243         skip_bits(&h->gb, 8);
 244
 245     return 0;
 246 }
 247
 248 static int decode_recovery_point(H264Context *h)
 249 {
 250     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
 251
 252     /* 1b exact_match_flag,
 253      * 1b broken_link_flag,
 254      * 2b changing_slice_group_idc */
 255     skip_bits(&h->gb, 4);
 256
 257     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 258         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
 259
 260     h->has_recovery_point = 1;
 261
 262     return 0;
 263 }
 264
 265 static int decode_buffering_period(H264Context *h)
 266 {
 267     unsigned int sps_id;
 268     int sched_sel_idx;
 269     SPS *sps;
 270
 271     sps_id = get_ue_golomb_31(&h->gb);
 272     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
 273         av_log(h->avctx, AV_LOG_ERROR,
 274                "non-existing SPS %d referenced in buffering period\n", sps_id);
 275         return AVERROR_INVALIDDATA;
 276     }
 277     sps = h->sps_buffers[sps_id];
 278
 279     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
 280     if (sps->nal_hrd_parameters_present_flag) {
 281         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 282             h->initial_cpb_removal_delay[sched_sel_idx] =
 283                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 284             // initial_cpb_removal_delay_offset
 285             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 286         }
 287     }
 288     if (sps->vcl_hrd_parameters_present_flag) {
 289         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 290             h->initial_cpb_removal_delay[sched_sel_idx] =
 291                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 292             // initial_cpb_removal_delay_offset
 293             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 294         }
 295     }
 296
 297     h->sei_buffering_period_present = 1;
 298     return 0;
 299 }
 300
 301 static int decode_frame_packing_arrangement(H264Context *h)
 302 {
 303     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
 304     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
 305     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
 306
 307     if (h->sei_frame_packing_present) {
 308         h->sei_fpa.frame_packing_arrangement_type =
 309         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
 310         h->sei_fpa.quincunx_sampling_flag         =
 311         h->quincunx_subsampling           = get_bits1(&h->gb);
 312         h->sei_fpa.content_interpretation_type    =
 313         h->content_interpretation_type    = get_bits(&h->gb, 6);
 314
 315         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
 316         // field_views_flag, current_frame_is_frame0_flag,
 317         // frame0_self_contained_flag, frame1_self_contained_flag
 318         skip_bits(&h->gb, 6);
 319
 320         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
 321             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
 322         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
 323         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
 324     }
 325     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
 326
 327     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 328         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
 329                                        h->sei_fpa.frame_packing_arrangement_id,
 330                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
 331                                        h->sei_fpa.frame_packing_arrangement_type,
 332                                        h->sei_fpa.quincunx_sampling_flag,
 333                                        h->sei_fpa.content_interpretation_type,
 334                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
 335
 336     return 0;
 337 }
 338
 339 static int decode_display_orientation(H264Context *h)
 340 {
 341     h->sei_display_orientation_present = !get_bits1(&h->gb);
 342
 343     if (h->sei_display_orientation_present) {
 344         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
 345         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
 346
 347         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
 348         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
 349         skip_bits1(&h->gb);     // display_orientation_extension_flag
 350     }
 351
 352     return 0;
 353 }
 354
 355 static int decode_GreenMetadata(H264Context *h)
 356 {
 357     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 358         av_log(h->avctx, AV_LOG_DEBUG,          "Green Metadata Info SEI message\n");
 359
 360     h->sei_green_metadata.green_metadata_type=get_bits(&h->gb, 8);
 361
 362     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 363         av_log(h->avctx, AV_LOG_DEBUG,          "green_metadata_type                            = %d\n",
 364                h->sei_green_metadata.green_metadata_type);
 365
 366     if (h->sei_green_metadata.green_metadata_type==0){
 367         h->sei_green_metadata.period_type=get_bits(&h->gb, 8);
 368
 369         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 370             av_log(h->avctx, AV_LOG_DEBUG,      "green_metadata_period_type                     = %d\n",
 371                    h->sei_green_metadata.period_type);
 372
 373         if (h->sei_green_metadata.green_metadata_type==2){
 374             h->sei_green_metadata.num_seconds = get_bits(&h->gb, 16);
 375             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 376                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_seconds                     = %d\n",
 377                        h->sei_green_metadata.num_seconds);
 378         }
 379         else if (h->sei_green_metadata.period_type==3){
 380             h->sei_green_metadata.num_pictures = get_bits(&h->gb, 16);
 381             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 382                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_pictures                    = %d\n",
 383                        h->sei_green_metadata.num_pictures);
 384         }
 385
 386         h->sei_green_metadata.percent_non_zero_macroblocks=get_bits(&h->gb, 8);
 387         h->sei_green_metadata.percent_intra_coded_macroblocks=get_bits(&h->gb, 8);
 388         h->sei_green_metadata.percent_six_tap_filtering=get_bits(&h->gb, 8);
 389         h->sei_green_metadata.percent_alpha_point_deblocking_instance=get_bits(&h->gb, 8);
 390
 391         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 392             av_log(h->avctx, AV_LOG_DEBUG,      "SEI GREEN Complexity Metrics                   = %f %f %f %f\n",
 393                                            (float)h->sei_green_metadata.percent_non_zero_macroblocks/255,
 394                                            (float)h->sei_green_metadata.percent_intra_coded_macroblocks/255,
 395                                            (float)h->sei_green_metadata.percent_six_tap_filtering/255,
 396                                            (float)h->sei_green_metadata.percent_alpha_point_deblocking_instance/255);
 397
 398     }else if( h->sei_green_metadata.green_metadata_type==1){
 399         h->sei_green_metadata.xsd_metric_type=get_bits(&h->gb, 8);
 400         h->sei_green_metadata.xsd_metric_value=get_bits(&h->gb, 16);
 401
 402         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 403             av_log(h->avctx, AV_LOG_DEBUG,      "xsd_metric_type                                = %d\n",
 404                    h->sei_green_metadata.xsd_metric_type);
 405         if ( h->sei_green_metadata.xsd_metric_type==0){
 406             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 407                 av_log(h->avctx, AV_LOG_DEBUG,  "xsd_metric_value                               = %f\n",
 408                        (float)h->sei_green_metadata.xsd_metric_value/100);
 409         }
 410     }
 411
 412     return 0;
 413 }
 414
 415 int ff_h264_decode_sei(H264Context *h)
 416 {
 417     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
 418         int type = 0;
 419         unsigned size = 0;
 420         unsigned next;
 421         int ret  = 0;
 422
 423         do {
 424             if (get_bits_left(&h->gb) < 8)
 425                 return AVERROR_INVALIDDATA;
 426             type += show_bits(&h->gb, 8);
 427         } while (get_bits(&h->gb, 8) == 255);
 428
 429         do {
 430             if (get_bits_left(&h->gb) < 8)
 431                 return AVERROR_INVALIDDATA;
 432             size += show_bits(&h->gb, 8);
 433         } while (get_bits(&h->gb, 8) == 255);
 434
 435         if (h->avctx->debug&FF_DEBUG_STARTCODE)
 436             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
 437
 438         if (size > get_bits_left(&h->gb) / 8) {
 439             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
 440                    type, 8*size, get_bits_left(&h->gb));
 441             return AVERROR_INVALIDDATA;
 442         }
 443         next = get_bits_count(&h->gb) + 8 * size;
 444
 445         switch (type) {
 446         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
 447             ret = decode_picture_timing(h);
 448             break;
 449         case SEI_TYPE_USER_DATA_REGISTERED:
 450             ret = decode_registered_user_data(h, size);
 451             break;
 452         case SEI_TYPE_USER_DATA_UNREGISTERED:
 453             ret = decode_unregistered_user_data(h, size);
 454             break;
 455         case SEI_TYPE_RECOVERY_POINT:
 456             ret = decode_recovery_point(h);
 457             break;
 458         case SEI_TYPE_BUFFERING_PERIOD:
 459             ret = decode_buffering_period(h);
 460             break;
 461         case SEI_TYPE_FRAME_PACKING:
 462             ret = decode_frame_packing_arrangement(h);
 463             break;
 464         case SEI_TYPE_DISPLAY_ORIENTATION:
 465             ret = decode_display_orientation(h);
 466             break;
 467         case SEI_TYPE_GREEN_METADATA:
 468             ret = decode_GreenMetadata(h);
 469             break;
 470         default:
 471             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
 472         }
 473         if (ret < 0)
 474             return ret;
 475
 476         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
 477
 478         // FIXME check bits here
 479         align_get_bits(&h->gb);
 480     }
 481
 482     return 0;
 483 }
 484
 485 const char* ff_h264_sei_stereo_mode(H264Context *h)
 486 {
 487     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
 488         switch (h->sei_fpa.frame_packing_arrangement_type) {
 489             case SEI_FPA_TYPE_CHECKERBOARD:
 490                 if (h->sei_fpa.content_interpretation_type == 2)
 491                     return "checkerboard_rl";
 492                 else
 493                     return "checkerboard_lr";
 494             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
 495                 if (h->sei_fpa.content_interpretation_type == 2)
 496                     return "col_interleaved_rl";
 497                 else
 498                     return "col_interleaved_lr";
 499             case SEI_FPA_TYPE_INTERLEAVE_ROW:
 500                 if (h->sei_fpa.content_interpretation_type == 2)
 501                     return "row_interleaved_rl";
 502                 else
 503                     return "row_interleaved_lr";
 504             case SEI_FPA_TYPE_SIDE_BY_SIDE:
 505                 if (h->sei_fpa.content_interpretation_type == 2)
 506                     return "right_left";
 507                 else
 508                     return "left_right";
 509             case SEI_FPA_TYPE_TOP_BOTTOM:
 510                 if (h->sei_fpa.content_interpretation_type == 2)
 511                     return "bottom_top";
 512                 else
 513                     return "top_bottom";
 514             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
 515                 if (h->sei_fpa.content_interpretation_type == 2)
 516                     return "block_rl";
 517                 else
 518                     return "block_lr";
 519             case SEI_FPA_TYPE_2D:
 520             default:
 521                 return "mono";
 522         }
 523     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
 524         return "mono";
 525     } else {
 526         return NULL;
 527     }
 528 }