codecs/hevc.h

   1 /*****************************************************************************
   2  * hevc.h:
   3  *****************************************************************************
   4  * Copyright (C) 2013-2014 L-SMASH project
   5  *
   6  * Authors: Yusuke Nakamura <muken.the.vfrmaniac@gmail.com>
   7  *
   8  * Permission to use, copy, modify, and/or distribute this software for any
   9  * purpose with or without fee is hereby granted, provided that the above
  10  * copyright notice and this permission notice appear in all copies.
  11  *
  12  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  13  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  14  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  15  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  16  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  17  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  18  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  19  *****************************************************************************/
  20
  21 /* This file is available under an ISC license. */
  22
  23 enum
  24 {
  25     HEVC_NALU_TYPE_TRAIL_N        = 0,
  26     HEVC_NALU_TYPE_TRAIL_R        = 1,
  27     HEVC_NALU_TYPE_TSA_N          = 2,
  28     HEVC_NALU_TYPE_TSA_R          = 3,
  29     HEVC_NALU_TYPE_STSA_N         = 4,
  30     HEVC_NALU_TYPE_STSA_R         = 5,
  31     HEVC_NALU_TYPE_RADL_N         = 6,
  32     HEVC_NALU_TYPE_RADL_R         = 7,
  33     HEVC_NALU_TYPE_RASL_N         = 8,
  34     HEVC_NALU_TYPE_RASL_R         = 9,
  35     HEVC_NALU_TYPE_RSV_VCL_R15    = 15,
  36     HEVC_NALU_TYPE_BLA_W_LP       = 16,
  37     HEVC_NALU_TYPE_BLA_W_RADL     = 17,
  38     HEVC_NALU_TYPE_BLA_N_LP       = 18,
  39     HEVC_NALU_TYPE_IDR_W_RADL     = 19,
  40     HEVC_NALU_TYPE_IDR_N_LP       = 20,
  41     HEVC_NALU_TYPE_CRA            = 21,
  42     HEVC_NALU_TYPE_RSV_IRAP_VCL22 = 22,
  43     HEVC_NALU_TYPE_RSV_IRAP_VCL23 = 23,
  44     HEVC_NALU_TYPE_RSV_VCL31      = 31,
  45     HEVC_NALU_TYPE_VPS            = 32,
  46     HEVC_NALU_TYPE_SPS            = 33,
  47     HEVC_NALU_TYPE_PPS            = 34,
  48     HEVC_NALU_TYPE_AUD            = 35,
  49     HEVC_NALU_TYPE_EOS            = 36,
  50     HEVC_NALU_TYPE_EOB            = 37,
  51     HEVC_NALU_TYPE_FD             = 38,
  52     HEVC_NALU_TYPE_PREFIX_SEI     = 39,
  53     HEVC_NALU_TYPE_SUFFIX_SEI     = 40,
  54     HEVC_NALU_TYPE_RSV_NVCL41     = 41,
  55     HEVC_NALU_TYPE_RSV_NVCL44     = 44,
  56     HEVC_NALU_TYPE_RSV_NVCL47     = 47,
  57     HEVC_NALU_TYPE_UNSPEC48       = 48,
  58     HEVC_NALU_TYPE_UNSPEC55       = 55,
  59     HEVC_NALU_TYPE_UNSPEC63       = 63,
  60     HEVC_NALU_TYPE_UNKNOWN        = 64
  61 };
  62
  63 typedef struct
  64 {
  65     uint8_t             array_completeness;
  66     uint8_t             NAL_unit_type;
  67     lsmash_entry_list_t list[1];
  68 } hevc_parameter_array_t;
  69
  70 struct lsmash_hevc_parameter_arrays_tag
  71 {
  72     hevc_parameter_array_t ps_array[HEVC_DCR_NALU_TYPE_NUM];
  73 };
  74
  75 typedef struct
  76 {
  77     unsigned forbidden_zero_bit : 1;
  78     unsigned nal_unit_type      : 7;   /* +1 bit for HEVC_NALU_TYPE_UNKNOWN */
  79     unsigned nuh_layer_id       : 6;
  80     unsigned TemporalId         : 3;
  81     unsigned length             : 15;
  82 } hevc_nalu_header_t;
  83
  84 /* Profile, Tier and Level */
  85 typedef struct
  86 {
  87     uint8_t  profile_space;
  88     uint8_t  tier_flag;
  89     uint8_t  profile_idc;
  90     uint32_t profile_compatibility_flags;
  91     uint8_t  progressive_source_flag;
  92     uint8_t  interlaced_source_flag;
  93     uint8_t  non_packed_constraint_flag;
  94     uint8_t  frame_only_constraint_flag;
  95     uint64_t reserved_zero_44bits;
  96     uint8_t  level_idc;
  97 } hevc_ptl_common_t;
  98
  99 typedef struct
 100 {
 101     hevc_ptl_common_t general;
 102     hevc_ptl_common_t sub_layer[6];
 103 } hevc_ptl_t;
 104
 105 /* HRD (Hypothetical Reference Decoder) */
 106 typedef struct
 107 {
 108     uint8_t  present;
 109     uint8_t  CpbDpbDelaysPresentFlag;
 110     uint8_t  sub_pic_hrd_params_present_flag;
 111     uint8_t  du_cpb_removal_delay_increment_length;
 112     uint8_t  sub_pic_cpb_params_in_pic_timing_sei_flag;
 113     uint8_t  dpb_output_delay_du_length;
 114     uint8_t  au_cpb_removal_delay_length;
 115     uint8_t  dpb_output_delay_length;
 116     uint8_t  fixed_pic_rate_general_flag[7];
 117     uint16_t elemental_duration_in_tc   [7];
 118 } hevc_hrd_t;
 119
 120 /* VPS (Video Parameter Set) */
 121 typedef struct
 122 {
 123     uint8_t    present;
 124     uint8_t    video_parameter_set_id;
 125     uint8_t    max_sub_layers_minus1;
 126     uint8_t    temporal_id_nesting_flag;
 127     uint8_t    timing_info_present_flag;
 128     uint8_t    frame_field_info_present_flag;
 129     uint16_t   num_hrd_parameters;
 130     hevc_ptl_t ptl;
 131     hevc_hrd_t hrd[2];
 132 } hevc_vps_t;
 133
 134 typedef struct
 135 {
 136     uint8_t       present;
 137     uint16_t      sar_width;
 138     uint16_t      sar_height;
 139     uint8_t       video_full_range_flag;
 140     uint8_t       colour_description_present_flag;
 141     uint8_t       colour_primaries;
 142     uint8_t       transfer_characteristics;
 143     uint8_t       matrix_coeffs;
 144     uint8_t       field_seq_flag;
 145     uint8_t       frame_field_info_present_flag;
 146     uint32_t      num_units_in_tick;
 147     uint32_t      time_scale;
 148     uint16_t      min_spatial_segmentation_idc;
 149     lsmash_crop_t def_disp_win_offset;
 150     hevc_hrd_t    hrd;
 151 } hevc_vui_t;
 152
 153 /* Short term reference picture sets */
 154 typedef struct
 155 {
 156     uint8_t NumNegativePics;
 157     uint8_t NumPositivePics;
 158     uint8_t NumDeltaPocs;
 159     uint8_t UsedByCurrPicS0[16];
 160     uint8_t UsedByCurrPicS1[16];
 161     int32_t DeltaPocS0[16];
 162     int32_t DeltaPocS1[16];
 163 } hevc_st_rps_t;
 164
 165 /* SPS (Sequence Parameter Set) */
 166 typedef struct
 167 {
 168     uint8_t       present;
 169     uint8_t       video_parameter_set_id;
 170     uint8_t       max_sub_layers_minus1;
 171     uint8_t       temporal_id_nesting_flag;
 172     hevc_ptl_t    ptl;
 173     uint8_t       seq_parameter_set_id;
 174     uint8_t       chroma_format_idc;
 175     uint8_t       separate_colour_plane_flag;
 176     uint8_t       bit_depth_luma_minus8;
 177     uint8_t       bit_depth_chroma_minus8;
 178     uint8_t       log2_max_pic_order_cnt_lsb;
 179     uint8_t       num_short_term_ref_pic_sets;
 180     uint8_t       long_term_ref_pics_present_flag;
 181     uint8_t       num_long_term_ref_pics_sps;
 182     uint8_t       temporal_mvp_enabled_flag;
 183     uint32_t      cropped_width;
 184     uint32_t      cropped_height;
 185     uint32_t      PicWidthInCtbsY;
 186     uint32_t      PicHeightInCtbsY;
 187     uint64_t      PicSizeInCtbsY;
 188     hevc_st_rps_t st_rps[65];
 189     hevc_vui_t    vui;
 190 } hevc_sps_t;
 191
 192 /* PPS (Picture Parameter Set) */
 193 typedef struct
 194 {
 195     uint8_t   present;
 196     uint8_t   pic_parameter_set_id;
 197     uint8_t   seq_parameter_set_id;
 198     uint8_t   dependent_slice_segments_enabled_flag;
 199     uint8_t   output_flag_present_flag;
 200     uint8_t   num_extra_slice_header_bits;
 201     uint8_t   tiles_enabled_flag;
 202     uint8_t   entropy_coding_sync_enabled_flag;
 203     uint32_t  num_tile_columns_minus1;
 204     uint32_t  num_tile_rows_minus1;
 205 #define SIZEOF_PPS_EXCLUDING_HEAP (sizeof(hevc_pps_t) - offsetof( hevc_pps_t, col_alloc_size ))
 206     size_t    col_alloc_size;
 207     size_t    row_alloc_size;
 208     uint32_t *colWidth;
 209     uint32_t *colBd;
 210     uint32_t *rowHeight;
 211     uint32_t *rowBd;
 212 } hevc_pps_t;
 213
 214 /* SEI (Supplemental Enhancement Information) */
 215 typedef struct
 216 {
 217     uint8_t present;
 218     uint8_t pic_struct;
 219 } hevc_pic_timing_t;
 220
 221 typedef struct
 222 {
 223     uint8_t present;
 224     uint8_t broken_link_flag;
 225     int32_t recovery_poc_cnt;
 226 } hevc_recovery_point_t;
 227
 228 typedef struct
 229 {
 230     hevc_pic_timing_t     pic_timing;
 231     hevc_recovery_point_t recovery_point;
 232 } hevc_sei_t;
 233
 234 /* Slice segment */
 235 typedef struct
 236 {
 237     uint8_t  present;
 238     uint8_t  nalu_type;
 239     uint8_t  TemporalId;
 240     uint8_t  type;
 241     uint8_t  video_parameter_set_id;
 242     uint8_t  seq_parameter_set_id;
 243     uint8_t  pic_parameter_set_id;
 244     uint8_t  first_slice_segment_in_pic_flag;
 245     uint8_t  dependent_slice_segment_flag;
 246     uint64_t segment_address;
 247     int32_t  pic_order_cnt_lsb;
 248 } hevc_slice_info_t;
 249
 250 /* Picture */
 251 typedef enum
 252 {
 253     HEVC_PICTURE_TYPE_I     = 0,
 254     HEVC_PICTURE_TYPE_I_P   = 1,
 255     HEVC_PICTURE_TYPE_I_P_B = 2,
 256     HEVC_PICTURE_TYPE_IDR   = 3,
 257     HEVC_PICTURE_TYPE_CRA   = 4,
 258     HEVC_PICTURE_TYPE_BLA   = 5,
 259     HEVC_PICTURE_TYPE_NONE  = 6,
 260 } hevc_picture_type;
 261
 262 typedef struct
 263 {
 264     hevc_picture_type type;
 265     uint8_t           irap;             /* 1: IDR, CRA or BLA picture */
 266     uint8_t           idr;              /* 1: IDR picture */
 267     uint8_t           broken_link;      /* 1: BLA picture or picture with broken link flag */
 268     uint8_t           radl;             /* 1: RADL picture */
 269     uint8_t           rasl;             /* 1: RASL picture */
 270     uint8_t           sublayer_nonref;  /* 1: sub-layer non-reference picture */
 271     uint8_t           closed_rap;       /* 1: no undecodable leading picture in CVS */
 272     uint8_t           random_accessible;/* 1: RAP or starting point of GDR */
 273     uint8_t           TemporalId;
 274     uint8_t           independent;      /* 1: intra coded picture */
 275     uint8_t           field_coded;      /* 1: field coded picture */
 276     uint8_t           pic_parameter_set_id;
 277     uint8_t           has_primary;      /* 1: an independent slice segment is present. */
 278     uint8_t           delta;
 279     /* POC */
 280     uint16_t          poc_lsb;
 281     int32_t           poc;
 282     int32_t           tid0_poc_msb;
 283     int32_t           tid0_poc_lsb;
 284     /* */
 285     int32_t           recovery_poc_cnt;
 286 } hevc_picture_info_t;
 287
 288 /* Access unit */
 289 typedef struct
 290 {
 291     uint8_t            *data;
 292     uint8_t            *incomplete_data;
 293     uint32_t            length;
 294     uint32_t            incomplete_length;
 295     uint32_t            number;
 296     uint8_t             TemporalId;
 297     hevc_picture_info_t picture;
 298 } hevc_access_unit_t;
 299
 300 typedef struct hevc_info_tag hevc_info_t;
 301
 302 typedef struct
 303 {
 304     lsmash_multiple_buffers_t *bank;
 305     uint8_t                   *rbsp;
 306 } hevc_stream_buffer_t;
 307
 308 struct hevc_info_tag
 309 {
 310     lsmash_hevc_specific_parameters_t hvcC_param;
 311     lsmash_hevc_specific_parameters_t hvcC_param_next;
 312     hevc_nalu_header_t   nuh;
 313     lsmash_entry_list_t  vps_list[1];
 314     lsmash_entry_list_t  sps_list[1];
 315     lsmash_entry_list_t  pps_list[1];
 316     hevc_vps_t           vps;           /* active VPS */
 317     hevc_sps_t           sps;           /* active SPS */
 318     hevc_pps_t           pps;           /* active PPS */
 319     hevc_sei_t           sei;           /* active SEI */
 320     hevc_slice_info_t    slice;         /* active slice */
 321     hevc_access_unit_t   au;
 322     uint8_t              prev_nalu_type;
 323     uint8_t              hvcC_pending;
 324     uint8_t              eos;           /* end of sequence */
 325     uint64_t             ebsp_head_pos;
 326     lsmash_bits_t       *bits;
 327     hevc_stream_buffer_t buffer;
 328 };
 329
 330 int hevc_setup_parser
 331 (
 332     hevc_info_t *info,
 333     int          parse_only
 334 );
 335
 336 void hevc_cleanup_parser
 337 (
 338     hevc_info_t *info
 339 );
 340
 341 uint64_t hevc_find_next_start_code
 342 (
 343     lsmash_bs_t        *bs,
 344     hevc_nalu_header_t *nuh,
 345     uint64_t           *start_code_length,
 346     uint64_t           *trailing_zero_bytes
 347 );
 348
 349 int hevc_calculate_poc
 350 (
 351     hevc_info_t         *info,
 352     hevc_picture_info_t *picture,
 353     hevc_picture_info_t *prev_picture
 354 );
 355
 356 void hevc_update_picture_info_for_slice
 357 (
 358     hevc_info_t         *info,
 359     hevc_picture_info_t *picture,
 360     hevc_slice_info_t   *slice
 361 );
 362
 363 void hevc_update_picture_info
 364 (
 365     hevc_info_t         *info,
 366     hevc_picture_info_t *picture,
 367     hevc_slice_info_t   *slice,
 368     hevc_sps_t          *sps,
 369     hevc_sei_t          *sei
 370 );
 371
 372 int hevc_find_au_delimit_by_slice_info
 373 (
 374     hevc_info_t       *info,
 375     hevc_slice_info_t *slice,
 376     hevc_slice_info_t *prev_slice
 377 );
 378
 379 int hevc_find_au_delimit_by_nalu_type
 380 (
 381     uint8_t nalu_type,
 382     uint8_t prev_nalu_type
 383 );
 384
 385 int hevc_supplement_buffer
 386 (
 387     hevc_stream_buffer_t *hb,
 388     hevc_access_unit_t   *au,
 389     uint32_t              size
 390 );
 391
 392 int hevc_parse_vps
 393 (
 394     hevc_info_t *info,
 395     uint8_t     *rbsp_buffer,
 396     uint8_t     *ebsp,
 397     uint64_t     ebsp_size
 398 );
 399
 400 int hevc_parse_sps
 401 (
 402     hevc_info_t *info,
 403     uint8_t     *rbsp_buffer,
 404     uint8_t     *ebsp,
 405     uint64_t     ebsp_size
 406 );
 407
 408 int hevc_parse_pps
 409 (
 410     hevc_info_t *info,
 411     uint8_t     *rbsp_buffer,
 412     uint8_t     *ebsp,
 413     uint64_t     ebsp_size
 414 );
 415
 416 int hevc_parse_sei
 417 (
 418     lsmash_bits_t      *bits,
 419     hevc_vps_t         *vps,
 420     hevc_sps_t         *sps,
 421     hevc_sei_t         *sei,
 422     hevc_nalu_header_t *nuh,
 423     uint8_t            *rbsp_buffer,
 424     uint8_t            *ebsp,
 425     uint64_t            ebsp_size
 426 );
 427 int hevc_parse_slice_segment_header
 428 (
 429     hevc_info_t        *info,
 430     hevc_nalu_header_t *nuh,
 431     uint8_t            *rbsp_buffer,
 432     uint8_t            *ebsp,
 433     uint64_t            ebsp_size
 434 );
 435
 436 int hevc_try_to_append_dcr_nalu
 437 (
 438     hevc_info_t              *info,
 439     lsmash_hevc_dcr_nalu_type ps_type,
 440     void                     *ps_data,
 441     uint32_t                  ps_length
 442 );
 443
 444 int hevc_move_pending_hvcC_param
 445 (
 446     hevc_info_t *info
 447 );