codecs/h264.h

   1 /*****************************************************************************
   2  * h264.h:
   3  *****************************************************************************
   4  * Copyright (C) 2012-2015 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     H264_NALU_TYPE_UNSPECIFIED0  = 0,   /* Unspecified */
  26     H264_NALU_TYPE_SLICE_N_IDR   = 1,   /* Coded slice of a non-IDR picture */
  27     H264_NALU_TYPE_SLICE_DP_A    = 2,   /* Coded slice data partition A */
  28     H264_NALU_TYPE_SLICE_DP_B    = 3,   /* Coded slice data partition B */
  29     H264_NALU_TYPE_SLICE_DP_C    = 4,   /* Coded slice data partition C */
  30     H264_NALU_TYPE_SLICE_IDR     = 5,   /* Coded slice of an IDR picture */
  31     H264_NALU_TYPE_SEI           = 6,   /* Supplemental Enhancement Information */
  32     H264_NALU_TYPE_SPS           = 7,   /* Sequence Parameter Set */
  33     H264_NALU_TYPE_PPS           = 8,   /* Picture Parameter Set */
  34     H264_NALU_TYPE_AUD           = 9,   /* Access Unit Delimiter */
  35     H264_NALU_TYPE_EOS           = 10,  /* End of Sequence */
  36     H264_NALU_TYPE_EOB           = 11,  /* End of Bitstream */
  37     H264_NALU_TYPE_FD            = 12,  /* Filler Data */
  38     H264_NALU_TYPE_SPS_EXT       = 13,  /* Sequence Parameter Set Extension */
  39     H264_NALU_TYPE_PREFIX        = 14,  /* Prefix NAL unit */
  40     H264_NALU_TYPE_SUBSET_SPS    = 15,  /* Subset Sequence Parameter Set */
  41     H264_NALU_TYPE_RSV_NVCL16    = 16,  /* Reserved */
  42     H264_NALU_TYPE_RSV_NVCL17    = 17,  /* Reserved */
  43     H264_NALU_TYPE_RSV_NVCL18    = 18,  /* Reserved */
  44     H264_NALU_TYPE_SLICE_AUX     = 19,  /* Coded slice of an auxiliary coded picture without partitioning */
  45     H264_NALU_TYPE_SLICE_EXT     = 20,  /* Coded slice extension */
  46     H264_NALU_TYPE_SLICE_EXT_DVC = 21,  /* Coded slice extension for depth view components */
  47     H264_NALU_TYPE_RSV22         = 22,  /* Reserved */
  48     H264_NALU_TYPE_RSV23         = 23,  /* Reserved */
  49     H264_NALU_TYPE_UNSPECIFIED24 = 24,  /* Unspecified */
  50     H264_NALU_TYPE_UNSPECIFIED31 = 31,  /* Unspecified */
  51 };
  52
  53 struct lsmash_h264_parameter_sets_tag
  54 {
  55     /* Each list contains entries as isom_dcr_ps_entry_t. */
  56     lsmash_entry_list_t sps_list   [1];
  57     lsmash_entry_list_t pps_list   [1];
  58     lsmash_entry_list_t spsext_list[1];
  59 };
  60
  61 typedef struct
  62 {
  63     unsigned forbidden_zero_bit : 1;
  64     unsigned nal_ref_idc        : 2;
  65     unsigned nal_unit_type      : 5;
  66     uint8_t  length;
  67 } h264_nalu_header_t;
  68
  69 typedef struct
  70 {
  71     uint8_t present;
  72     uint8_t CpbDpbDelaysPresentFlag;
  73     uint8_t cpb_removal_delay_length;
  74     uint8_t dpb_output_delay_length;
  75 } h264_hrd_t;
  76
  77 typedef struct
  78 {
  79     uint16_t sar_width;
  80     uint16_t sar_height;
  81     uint8_t  video_full_range_flag;
  82     uint8_t  colour_primaries;
  83     uint8_t  transfer_characteristics;
  84     uint8_t  matrix_coefficients;
  85     uint32_t num_units_in_tick;
  86     uint32_t time_scale;
  87     uint8_t  fixed_frame_rate_flag;
  88     uint8_t  pic_struct_present_flag;
  89     h264_hrd_t hrd;
  90 } h264_vui_t;
  91
  92 typedef struct
  93 {
  94     uint8_t  present;
  95     uint8_t  profile_idc;
  96     uint8_t  constraint_set_flags;
  97     uint8_t  level_idc;
  98     uint8_t  seq_parameter_set_id;
  99     uint8_t  chroma_format_idc;
 100     uint8_t  separate_colour_plane_flag;
 101     uint8_t  ChromaArrayType;
 102     uint8_t  bit_depth_luma_minus8;
 103     uint8_t  bit_depth_chroma_minus8;
 104     uint8_t  log2_max_frame_num;
 105     uint8_t  pic_order_cnt_type;
 106     uint8_t  delta_pic_order_always_zero_flag;
 107     uint8_t  num_ref_frames_in_pic_order_cnt_cycle;
 108     uint8_t  frame_mbs_only_flag;
 109     int32_t  offset_for_non_ref_pic;
 110     int32_t  offset_for_top_to_bottom_field;
 111     int32_t  offset_for_ref_frame[255];
 112     int64_t  ExpectedDeltaPerPicOrderCntCycle;
 113     uint32_t max_num_ref_frames;
 114     uint32_t MaxFrameNum;
 115     uint32_t log2_max_pic_order_cnt_lsb;
 116     uint32_t MaxPicOrderCntLsb;
 117     uint32_t PicSizeInMapUnits;
 118     uint32_t cropped_width;
 119     uint32_t cropped_height;
 120     h264_vui_t vui;
 121 } h264_sps_t;
 122
 123 typedef struct
 124 {
 125     uint8_t  present;
 126     uint8_t  pic_parameter_set_id;
 127     uint8_t  seq_parameter_set_id;
 128     uint8_t  entropy_coding_mode_flag;
 129     uint8_t  bottom_field_pic_order_in_frame_present_flag;
 130     uint8_t  num_slice_groups_minus1;
 131     uint8_t  slice_group_map_type;
 132     uint8_t  num_ref_idx_l0_default_active_minus1;
 133     uint8_t  num_ref_idx_l1_default_active_minus1;
 134     uint8_t  weighted_pred_flag;
 135     uint8_t  weighted_bipred_idc;
 136     uint8_t  deblocking_filter_control_present_flag;
 137     uint8_t  redundant_pic_cnt_present_flag;
 138     uint32_t SliceGroupChangeRate;
 139 } h264_pps_t;
 140
 141 typedef struct
 142 {
 143     uint8_t present;
 144     uint8_t pic_struct;
 145 } h264_pic_timing_t;
 146
 147 typedef struct
 148 {
 149     uint8_t  present;
 150     uint8_t  random_accessible;
 151     uint8_t  broken_link_flag;
 152     uint32_t recovery_frame_cnt;
 153 } h264_recovery_point_t;
 154
 155 typedef struct
 156 {
 157     h264_pic_timing_t     pic_timing;
 158     h264_recovery_point_t recovery_point;
 159 } h264_sei_t;
 160
 161 typedef struct
 162 {
 163     uint8_t  present;
 164     uint8_t  slice_id;  /* only for slice data partition */
 165     uint8_t  type;
 166     uint8_t  pic_order_cnt_type;
 167     uint8_t  nal_ref_idc;
 168     uint8_t  IdrPicFlag;
 169     uint8_t  seq_parameter_set_id;
 170     uint8_t  pic_parameter_set_id;
 171     uint8_t  field_pic_flag;
 172     uint8_t  bottom_field_flag;
 173     uint8_t  has_mmco5;
 174     uint8_t  has_redundancy;
 175     uint16_t idr_pic_id;
 176     uint32_t frame_num;
 177     int32_t  pic_order_cnt_lsb;
 178     int32_t  delta_pic_order_cnt_bottom;
 179     int32_t  delta_pic_order_cnt[2];
 180 } h264_slice_info_t;
 181
 182 typedef enum
 183 {
 184     H264_PICTURE_TYPE_IDR         = 0,
 185     H264_PICTURE_TYPE_I           = 1,
 186     H264_PICTURE_TYPE_I_P         = 2,
 187     H264_PICTURE_TYPE_I_P_B       = 3,
 188     H264_PICTURE_TYPE_SI          = 4,
 189     H264_PICTURE_TYPE_SI_SP       = 5,
 190     H264_PICTURE_TYPE_I_SI        = 6,
 191     H264_PICTURE_TYPE_I_SI_P_SP   = 7,
 192     H264_PICTURE_TYPE_I_SI_P_SP_B = 8,
 193     H264_PICTURE_TYPE_NONE        = 9,
 194 } h264_picture_type;
 195
 196 typedef struct
 197 {
 198     h264_picture_type type;
 199     uint8_t           idr;
 200     uint8_t           random_accessible;
 201     uint8_t           independent;
 202     uint8_t           disposable;   /* 1: nal_ref_idc == 0, 0: otherwise */
 203     uint8_t           has_redundancy;
 204     uint8_t           has_primary;
 205     uint8_t           pic_parameter_set_id;
 206     uint8_t           field_pic_flag;
 207     uint8_t           bottom_field_flag;
 208     uint8_t           delta;
 209     uint8_t           broken_link_flag;
 210     /* POC */
 211     uint8_t           has_mmco5;
 212     uint8_t           ref_pic_has_mmco5;
 213     uint8_t           ref_pic_bottom_field_flag;
 214     int32_t           ref_pic_TopFieldOrderCnt;
 215     int32_t           ref_pic_PicOrderCntMsb;
 216     int32_t           ref_pic_PicOrderCntLsb;
 217     int32_t           pic_order_cnt_lsb;
 218     int32_t           delta_pic_order_cnt_bottom;
 219     int32_t           delta_pic_order_cnt[2];
 220     int32_t           PicOrderCnt;
 221     uint32_t          FrameNumOffset;
 222     /* */
 223     uint32_t          recovery_frame_cnt;
 224     uint32_t          frame_num;
 225 } h264_picture_info_t;
 226
 227 typedef struct
 228 {
 229     uint8_t            *data;
 230     uint8_t            *incomplete_data;
 231     uint32_t            length;
 232     uint32_t            incomplete_length;
 233     uint32_t            number;
 234     h264_picture_info_t picture;
 235 } h264_access_unit_t;
 236
 237 typedef struct h264_info_tag h264_info_t;
 238
 239 typedef struct
 240 {
 241     lsmash_multiple_buffers_t *bank;
 242     uint8_t                   *rbsp;
 243 } h264_stream_buffer_t;
 244
 245 struct h264_info_tag
 246 {
 247     lsmash_h264_specific_parameters_t avcC_param;
 248     lsmash_h264_specific_parameters_t avcC_param_next;
 249     lsmash_entry_list_t  sps_list  [1]; /* contains entries as h264_sps_t */
 250     lsmash_entry_list_t  pps_list  [1]; /* contains entries as h264_pps_t */
 251     lsmash_entry_list_t  slice_list[1]; /* for slice data partition */
 252     h264_sps_t           sps;           /* active SPS */
 253     h264_pps_t           pps;           /* active PPS */
 254     h264_sei_t           sei;           /* active SEI */
 255     h264_slice_info_t    slice;         /* active slice */
 256     h264_access_unit_t   au;
 257     uint8_t              prev_nalu_type;
 258     uint8_t              avcC_pending;
 259     lsmash_bits_t       *bits;
 260     h264_stream_buffer_t buffer;
 261 };
 262
 263 int h264_setup_parser
 264 (
 265     h264_info_t *info,
 266     int          parse_only
 267 );
 268
 269 void h264_cleanup_parser
 270 (
 271     h264_info_t *info
 272 );
 273
 274 uint64_t h264_find_next_start_code
 275 (
 276     lsmash_bs_t        *bs,
 277     h264_nalu_header_t *nuh,
 278     uint64_t           *start_code_length,
 279     uint64_t           *trailing_zero_bytes
 280 );
 281
 282 int h264_calculate_poc
 283 (
 284     h264_info_t         *info,
 285     h264_picture_info_t *picture,
 286     h264_picture_info_t *prev_picture
 287 );
 288
 289 void h264_update_picture_info_for_slice
 290 (
 291     h264_info_t         *info,
 292     h264_picture_info_t *picture,
 293     h264_slice_info_t   *slice
 294 );
 295
 296 void h264_update_picture_info
 297 (
 298     h264_info_t         *info,
 299     h264_picture_info_t *picture,
 300     h264_slice_info_t   *slice,
 301     h264_sei_t          *sei
 302 );
 303
 304 int h264_find_au_delimit_by_slice_info
 305 (
 306     h264_slice_info_t *slice,
 307     h264_slice_info_t *prev_slice
 308 );
 309
 310 int h264_find_au_delimit_by_nalu_type
 311 (
 312     uint8_t nalu_type,
 313     uint8_t prev_nalu_type
 314 );
 315
 316 int h264_supplement_buffer
 317 (
 318     h264_stream_buffer_t *buffer,
 319     h264_access_unit_t   *au,
 320     uint32_t              size
 321 );
 322
 323 int h264_parse_sps
 324 (
 325     h264_info_t *info,
 326     uint8_t     *rbsp_buffer,
 327     uint8_t     *ebsp,
 328     uint64_t     ebsp_size
 329 );
 330
 331 int h264_parse_pps
 332 (
 333     h264_info_t *info,
 334     uint8_t     *rbsp_buffer,
 335     uint8_t     *ebsp,
 336     uint64_t     ebsp_size
 337 );
 338
 339 int h264_parse_sei
 340 (
 341     lsmash_bits_t *bits,
 342     h264_sps_t    *sps,
 343     h264_sei_t    *sei,
 344     uint8_t       *rbsp_buffer,
 345     uint8_t       *ebsp,
 346     uint64_t       ebsp_size
 347 );
 348
 349 int h264_parse_slice
 350 (
 351     h264_info_t        *info,
 352     h264_nalu_header_t *nuh,
 353     uint8_t            *rbsp_buffer,
 354     uint8_t            *ebsp,
 355     uint64_t            ebsp_size
 356 );
 357
 358 int h264_try_to_append_parameter_set
 359 (
 360     h264_info_t                   *info,
 361     lsmash_h264_parameter_set_type ps_type,
 362     void                          *ps_data,
 363     uint32_t                       ps_length
 364 );
 365
 366 int h264_move_pending_avcC_param
 367 (
 368     h264_info_t *info
 369 );