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(H264Context *h, int size)
 116 {
 117     uint32_t country_code;
 118     uint32_t user_identifier;
 119     int flag, cc_count, user_data_type_code;
 120
 121     if (size < 7)
 122         return AVERROR_INVALIDDATA;
 123     size -= 7;
 124
 125     country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
 126     if (country_code == 0xFF) {
 127         skip_bits(&h->gb, 8);           // itu_t_t35_country_code_extension_byte
 128         size--;
 129     }
 130
 131     /* itu_t_t35_payload_byte follows */
 132     skip_bits(&h->gb, 8);              // terminal provider code
 133     skip_bits(&h->gb, 8);              // terminal provider oriented code
 134     user_identifier = get_bits_long(&h->gb, 32);
 135
 136     switch (user_identifier) {
 137         case MKBETAG('D', 'T', 'G', '1'):       // afd_data
 138             if (size-- < 1)
 139                 return AVERROR_INVALIDDATA;
 140             skip_bits(&h->gb, 1);               // 0
 141             flag = get_bits(&h->gb, 1);         // active_format_flag
 142             skip_bits(&h->gb, 6);               // reserved
 143
 144             if (flag) {
 145                 if (size-- < 1)
 146                     return AVERROR_INVALIDDATA;
 147                 skip_bits(&h->gb, 4);           // reserved
 148                 h->active_format_description   = get_bits(&h->gb, 4);
 149                 h->sei_reguserdata_afd_present = 1;
 150 #if FF_API_AFD
 151 FF_DISABLE_DEPRECATION_WARNINGS
 152                 h->avctx->dtg_active_format = h->active_format_description;
 153 FF_ENABLE_DEPRECATION_WARNINGS
 154 #endif /* FF_API_AFD */
 155             }
 156             break;
 157         case MKBETAG('G', 'A', '9', '4'): // "GA94" closed captions
 158             if (size < 3)
 159                 return AVERROR(EINVAL);
 160             user_data_type_code = get_bits(&h->gb, 8);
 161             if (user_data_type_code == 0x3) {
 162                 skip_bits(&h->gb, 1);           // reserved
 163
 164                 flag = get_bits(&h->gb, 1);     // process_cc_data_flag
 165                 if (flag) {
 166                     skip_bits(&h->gb, 1);       // zero bit
 167                     cc_count = get_bits(&h->gb, 5);
 168                     skip_bits(&h->gb, 8);       // reserved
 169                     size -= 2;
 170
 171                     if (cc_count && size >= cc_count * 3) {
 172                         int i;
 173                         uint8_t *tmp;
 174                         if ((int64_t)h->a53_caption_size + (int64_t)cc_count*3 > INT_MAX)
 175                             return AVERROR(EINVAL);
 176
 177                         // Allow merging of the cc data from two fields
 178                         tmp = av_realloc(h->a53_caption, h->a53_caption_size + cc_count*3);
 179                         if (!tmp)
 180                             return AVERROR(ENOMEM);
 181                         h->a53_caption = tmp;
 182
 183                         for (i = 0; i < cc_count; i++) {
 184                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 185                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 186                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
 187                         }
 188
 189                         skip_bits(&h->gb, 8);   // marker_bits
 190                     }
 191                 }
 192             }
 193             break;
 194         default:
 195             skip_bits(&h->gb, size * 8);
 196             break;
 197     }
 198
 199     return 0;
 200 }
 201
 202 static int decode_unregistered_user_data(H264Context *h, int size)
 203 {
 204     uint8_t user_data[16 + 256];
 205     int e, build, i;
 206
 207     if (size < 16)
 208         return AVERROR_INVALIDDATA;
 209
 210     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
 211         user_data[i] = get_bits(&h->gb, 8);
 212
 213     user_data[i] = 0;
 214     e = sscanf(user_data + 16, "x264 - core %d", &build);
 215     if (e == 1 && build > 0)
 216         h->x264_build = build;
 217     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
 218         h->x264_build = 67;
 219
 220     if (h->avctx->debug & FF_DEBUG_BUGS)
 221         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
 222
 223     for (; i < size; i++)
 224         skip_bits(&h->gb, 8);
 225
 226     return 0;
 227 }
 228
 229 static int decode_recovery_point(H264Context *h)
 230 {
 231     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
 232
 233     /* 1b exact_match_flag,
 234      * 1b broken_link_flag,
 235      * 2b changing_slice_group_idc */
 236     skip_bits(&h->gb, 4);
 237
 238     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 239         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
 240
 241     h->has_recovery_point = 1;
 242
 243     return 0;
 244 }
 245
 246 static int decode_buffering_period(H264Context *h)
 247 {
 248     unsigned int sps_id;
 249     int sched_sel_idx;
 250     SPS *sps;
 251
 252     sps_id = get_ue_golomb_31(&h->gb);
 253     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
 254         av_log(h->avctx, AV_LOG_ERROR,
 255                "non-existing SPS %d referenced in buffering period\n", sps_id);
 256         return AVERROR_INVALIDDATA;
 257     }
 258     sps = h->sps_buffers[sps_id];
 259
 260     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
 261     if (sps->nal_hrd_parameters_present_flag) {
 262         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 263             h->initial_cpb_removal_delay[sched_sel_idx] =
 264                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 265             // initial_cpb_removal_delay_offset
 266             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 267         }
 268     }
 269     if (sps->vcl_hrd_parameters_present_flag) {
 270         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
 271             h->initial_cpb_removal_delay[sched_sel_idx] =
 272                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
 273             // initial_cpb_removal_delay_offset
 274             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
 275         }
 276     }
 277
 278     h->sei_buffering_period_present = 1;
 279     return 0;
 280 }
 281
 282 static int decode_frame_packing_arrangement(H264Context *h)
 283 {
 284     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
 285     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
 286     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
 287
 288     if (h->sei_frame_packing_present) {
 289         h->sei_fpa.frame_packing_arrangement_type =
 290         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
 291         h->sei_fpa.quincunx_sampling_flag         =
 292         h->quincunx_subsampling           = get_bits1(&h->gb);
 293         h->sei_fpa.content_interpretation_type    =
 294         h->content_interpretation_type    = get_bits(&h->gb, 6);
 295
 296         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
 297         // field_views_flag, current_frame_is_frame0_flag,
 298         // frame0_self_contained_flag, frame1_self_contained_flag
 299         skip_bits(&h->gb, 6);
 300
 301         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
 302             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
 303         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
 304         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
 305     }
 306     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
 307
 308     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
 309         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
 310                                        h->sei_fpa.frame_packing_arrangement_id,
 311                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
 312                                        h->sei_fpa.frame_packing_arrangement_type,
 313                                        h->sei_fpa.quincunx_sampling_flag,
 314                                        h->sei_fpa.content_interpretation_type,
 315                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
 316
 317     return 0;
 318 }
 319
 320 static int decode_display_orientation(H264Context *h)
 321 {
 322     h->sei_display_orientation_present = !get_bits1(&h->gb);
 323
 324     if (h->sei_display_orientation_present) {
 325         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
 326         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
 327
 328         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
 329         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
 330         skip_bits1(&h->gb);     // display_orientation_extension_flag
 331     }
 332
 333     return 0;
 334 }
 335
 336 static int decode_GreenMetadata(H264Context *h)
 337 {
 338     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 339         av_log(h->avctx, AV_LOG_DEBUG,          "Green Metadata Info SEI message\n");
 340
 341     h->sei_green_metadata.green_metadata_type=get_bits(&h->gb, 8);
 342
 343     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 344         av_log(h->avctx, AV_LOG_DEBUG,          "green_metadata_type                            = %d\n",
 345                h->sei_green_metadata.green_metadata_type);
 346
 347     if (h->sei_green_metadata.green_metadata_type==0){
 348         h->sei_green_metadata.period_type=get_bits(&h->gb, 8);
 349
 350         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 351             av_log(h->avctx, AV_LOG_DEBUG,      "green_metadata_period_type                     = %d\n",
 352                    h->sei_green_metadata.period_type);
 353
 354         if (h->sei_green_metadata.green_metadata_type==2){
 355             h->sei_green_metadata.num_seconds = get_bits(&h->gb, 16);
 356             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 357                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_seconds                     = %d\n",
 358                        h->sei_green_metadata.num_seconds);
 359         }
 360         else if (h->sei_green_metadata.period_type==3){
 361             h->sei_green_metadata.num_pictures = get_bits(&h->gb, 16);
 362             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 363                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_pictures                    = %d\n",
 364                        h->sei_green_metadata.num_pictures);
 365         }
 366
 367         h->sei_green_metadata.percent_non_zero_macroblocks=get_bits(&h->gb, 8);
 368         h->sei_green_metadata.percent_intra_coded_macroblocks=get_bits(&h->gb, 8);
 369         h->sei_green_metadata.percent_six_tap_filtering=get_bits(&h->gb, 8);
 370         h->sei_green_metadata.percent_alpha_point_deblocking_instance=get_bits(&h->gb, 8);
 371
 372         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 373             av_log(h->avctx, AV_LOG_DEBUG,      "SEI GREEN Complexity Metrics                   = %f %f %f %f\n",
 374                                            (float)h->sei_green_metadata.percent_non_zero_macroblocks/255,
 375                                            (float)h->sei_green_metadata.percent_intra_coded_macroblocks/255,
 376                                            (float)h->sei_green_metadata.percent_six_tap_filtering/255,
 377                                            (float)h->sei_green_metadata.percent_alpha_point_deblocking_instance/255);
 378
 379     }else if( h->sei_green_metadata.green_metadata_type==1){
 380         h->sei_green_metadata.xsd_metric_type=get_bits(&h->gb, 8);
 381         h->sei_green_metadata.xsd_metric_value=get_bits(&h->gb, 16);
 382
 383         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 384             av_log(h->avctx, AV_LOG_DEBUG,      "xsd_metric_type                                = %d\n",
 385                    h->sei_green_metadata.xsd_metric_type);
 386         if ( h->sei_green_metadata.xsd_metric_type==0){
 387             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
 388                 av_log(h->avctx, AV_LOG_DEBUG,  "xsd_metric_value                               = %f\n",
 389                        (float)h->sei_green_metadata.xsd_metric_value/100);
 390         }
 391     }
 392
 393     return 0;
 394 }
 395
 396 int ff_h264_decode_sei(H264Context *h)
 397 {
 398     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
 399         int type = 0;
 400         unsigned size = 0;
 401         unsigned next;
 402         int ret  = 0;
 403
 404         do {
 405             if (get_bits_left(&h->gb) < 8)
 406                 return AVERROR_INVALIDDATA;
 407             type += show_bits(&h->gb, 8);
 408         } while (get_bits(&h->gb, 8) == 255);
 409
 410         do {
 411             if (get_bits_left(&h->gb) < 8)
 412                 return AVERROR_INVALIDDATA;
 413             size += show_bits(&h->gb, 8);
 414         } while (get_bits(&h->gb, 8) == 255);
 415
 416         if (h->avctx->debug&FF_DEBUG_STARTCODE)
 417             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
 418
 419         if (size > get_bits_left(&h->gb) / 8) {
 420             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
 421                    type, 8*size, get_bits_left(&h->gb));
 422             return AVERROR_INVALIDDATA;
 423         }
 424         next = get_bits_count(&h->gb) + 8 * size;
 425
 426         switch (type) {
 427         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
 428             ret = decode_picture_timing(h);
 429             break;
 430         case SEI_TYPE_USER_DATA_REGISTERED:
 431             ret = decode_registered_user_data(h, size);
 432             break;
 433         case SEI_TYPE_USER_DATA_UNREGISTERED:
 434             ret = decode_unregistered_user_data(h, size);
 435             break;
 436         case SEI_TYPE_RECOVERY_POINT:
 437             ret = decode_recovery_point(h);
 438             break;
 439         case SEI_TYPE_BUFFERING_PERIOD:
 440             ret = decode_buffering_period(h);
 441             break;
 442         case SEI_TYPE_FRAME_PACKING:
 443             ret = decode_frame_packing_arrangement(h);
 444             break;
 445         case SEI_TYPE_DISPLAY_ORIENTATION:
 446             ret = decode_display_orientation(h);
 447             break;
 448         case SEI_TYPE_GREEN_METADATA:
 449             ret = decode_GreenMetadata(h);
 450             break;
 451         default:
 452             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
 453         }
 454         if (ret < 0)
 455             return ret;
 456
 457         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
 458
 459         // FIXME check bits here
 460         align_get_bits(&h->gb);
 461     }
 462
 463     return 0;
 464 }
 465
 466 const char* ff_h264_sei_stereo_mode(H264Context *h)
 467 {
 468     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
 469         switch (h->sei_fpa.frame_packing_arrangement_type) {
 470             case SEI_FPA_TYPE_CHECKERBOARD:
 471                 if (h->sei_fpa.content_interpretation_type == 2)
 472                     return "checkerboard_rl";
 473                 else
 474                     return "checkerboard_lr";
 475             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
 476                 if (h->sei_fpa.content_interpretation_type == 2)
 477                     return "col_interleaved_rl";
 478                 else
 479                     return "col_interleaved_lr";
 480             case SEI_FPA_TYPE_INTERLEAVE_ROW:
 481                 if (h->sei_fpa.content_interpretation_type == 2)
 482                     return "row_interleaved_rl";
 483                 else
 484                     return "row_interleaved_lr";
 485             case SEI_FPA_TYPE_SIDE_BY_SIDE:
 486                 if (h->sei_fpa.content_interpretation_type == 2)
 487                     return "right_left";
 488                 else
 489                     return "left_right";
 490             case SEI_FPA_TYPE_TOP_BOTTOM:
 491                 if (h->sei_fpa.content_interpretation_type == 2)
 492                     return "bottom_top";
 493                 else
 494                     return "top_bottom";
 495             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
 496                 if (h->sei_fpa.content_interpretation_type == 2)
 497                     return "block_rl";
 498                 else
 499                     return "block_lr";
 500             case SEI_FPA_TYPE_2D:
 501             default:
 502                 return "mono";
 503         }
 504     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
 505         return "mono";
 506     } else {
 507         return NULL;
 508     }
 509 }