git.sesse.net Git - ffmpeg/blob - libavcodec/cbs_h264.h

   1 /*
   2  * This file is part of FFmpeg.
   3  *
   4  * FFmpeg is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU Lesser General Public
   6  * License as published by the Free Software Foundation; either
   7  * version 2.1 of the License, or (at your option) any later version.
   8  *
   9  * FFmpeg is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12  * Lesser General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU Lesser General Public
  15  * License along with FFmpeg; if not, write to the Free Software
  16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  17  */
  18
  19 #ifndef AVCODEC_CBS_H264_H
  20 #define AVCODEC_CBS_H264_H
  21
  22 #include <stddef.h>
  23 #include <stdint.h>
  24
  25 #include "cbs.h"
  26 #include "cbs_h2645.h"
  27 #include "h264.h"
  28
  29
  30 enum {
  31     // This limit is arbitrary - it is sufficient for one message of each
  32     // type plus some repeats, and will therefore easily cover all sane
  33     // streams.  However, it is possible to make technically-valid streams
  34     // for which it will fail (for example, by including a large number of
  35     // user-data-unregistered messages).
  36     H264_MAX_SEI_PAYLOADS = 64,
  37 };
  38
  39
  40 typedef struct H264RawNALUnitHeader {
  41     uint8_t forbidden_zero_bit;
  42     uint8_t nal_ref_idc;
  43     uint8_t nal_unit_type;
  44
  45     uint8_t svc_extension_flag;
  46     uint8_t avc_3d_extension_flag;
  47 } H264RawNALUnitHeader;
  48
  49 typedef struct H264RawScalingList {
  50     int8_t delta_scale[64];
  51 } H264RawScalingList;
  52
  53 typedef struct H264RawHRD {
  54     uint8_t cpb_cnt_minus1;
  55     uint8_t bit_rate_scale;
  56     uint8_t cpb_size_scale;
  57
  58     uint32_t bit_rate_value_minus1[H264_MAX_CPB_CNT];
  59     uint32_t cpb_size_value_minus1[H264_MAX_CPB_CNT];
  60     uint8_t cbr_flag[H264_MAX_CPB_CNT];
  61
  62     uint8_t initial_cpb_removal_delay_length_minus1;
  63     uint8_t cpb_removal_delay_length_minus1;
  64     uint8_t dpb_output_delay_length_minus1;
  65     uint8_t time_offset_length;
  66 } H264RawHRD;
  67
  68 typedef struct H264RawVUI {
  69     uint8_t aspect_ratio_info_present_flag;
  70     uint8_t aspect_ratio_idc;
  71     uint16_t sar_width;
  72     uint16_t sar_height;
  73
  74     uint8_t overscan_info_present_flag;
  75     uint8_t overscan_appropriate_flag;
  76
  77     uint8_t video_signal_type_present_flag;
  78     uint8_t video_format;
  79     uint8_t video_full_range_flag;
  80     uint8_t colour_description_present_flag;
  81     uint8_t colour_primaries;
  82     uint8_t transfer_characteristics;
  83     uint8_t matrix_coefficients;
  84
  85     uint8_t chroma_loc_info_present_flag;
  86     uint8_t chroma_sample_loc_type_top_field;
  87     uint8_t chroma_sample_loc_type_bottom_field;
  88
  89     uint8_t timing_info_present_flag;
  90     uint32_t num_units_in_tick;
  91     uint32_t time_scale;
  92     uint8_t fixed_frame_rate_flag;
  93
  94     uint8_t nal_hrd_parameters_present_flag;
  95     H264RawHRD nal_hrd_parameters;
  96     uint8_t vcl_hrd_parameters_present_flag;
  97     H264RawHRD vcl_hrd_parameters;
  98     uint8_t low_delay_hrd_flag;
  99
 100     uint8_t pic_struct_present_flag;
 101
 102     uint8_t bitstream_restriction_flag;
 103     uint8_t motion_vectors_over_pic_boundaries_flag;
 104     uint8_t max_bytes_per_pic_denom;
 105     uint8_t max_bits_per_mb_denom;
 106     uint8_t log2_max_mv_length_horizontal;
 107     uint8_t log2_max_mv_length_vertical;
 108     uint8_t max_num_reorder_frames;
 109     uint8_t max_dec_frame_buffering;
 110 } H264RawVUI;
 111
 112 typedef struct H264RawSPS {
 113     H264RawNALUnitHeader nal_unit_header;
 114
 115     uint8_t profile_idc;
 116     uint8_t constraint_set0_flag;
 117     uint8_t constraint_set1_flag;
 118     uint8_t constraint_set2_flag;
 119     uint8_t constraint_set3_flag;
 120     uint8_t constraint_set4_flag;
 121     uint8_t constraint_set5_flag;
 122     uint8_t reserved_zero_2bits;
 123     uint8_t level_idc;
 124
 125     uint8_t seq_parameter_set_id;
 126
 127     uint8_t chroma_format_idc;
 128     uint8_t separate_colour_plane_flag;
 129     uint8_t bit_depth_luma_minus8;
 130     uint8_t bit_depth_chroma_minus8;
 131     uint8_t qpprime_y_zero_transform_bypass_flag;
 132
 133     uint8_t seq_scaling_matrix_present_flag;
 134     uint8_t seq_scaling_list_present_flag[12];
 135     H264RawScalingList scaling_list_4x4[6];
 136     H264RawScalingList scaling_list_8x8[6];
 137
 138     uint8_t log2_max_frame_num_minus4;
 139     uint8_t pic_order_cnt_type;
 140     uint8_t log2_max_pic_order_cnt_lsb_minus4;
 141     uint8_t delta_pic_order_always_zero_flag;
 142     int32_t offset_for_non_ref_pic;
 143     int32_t offset_for_top_to_bottom_field;
 144     uint8_t num_ref_frames_in_pic_order_cnt_cycle;
 145     int32_t offset_for_ref_frame[256];
 146
 147     uint8_t max_num_ref_frames;
 148     uint8_t gaps_in_frame_num_allowed_flag;
 149
 150     uint16_t pic_width_in_mbs_minus1;
 151     uint16_t pic_height_in_map_units_minus1;
 152
 153     uint8_t frame_mbs_only_flag;
 154     uint8_t mb_adaptive_frame_field_flag;
 155     uint8_t direct_8x8_inference_flag;
 156
 157     uint8_t frame_cropping_flag;
 158     uint16_t frame_crop_left_offset;
 159     uint16_t frame_crop_right_offset;
 160     uint16_t frame_crop_top_offset;
 161     uint16_t frame_crop_bottom_offset;
 162
 163     uint8_t vui_parameters_present_flag;
 164     H264RawVUI vui;
 165 } H264RawSPS;
 166
 167 typedef struct H264RawSPSExtension {
 168     H264RawNALUnitHeader nal_unit_header;
 169
 170     uint8_t seq_parameter_set_id;
 171
 172     uint8_t aux_format_idc;
 173     uint8_t bit_depth_aux_minus8;
 174     uint8_t alpha_incr_flag;
 175     uint16_t alpha_opaque_value;
 176     uint16_t alpha_transparent_value;
 177
 178     uint8_t additional_extension_flag;
 179 } H264RawSPSExtension;
 180
 181 typedef struct H264RawPPS {
 182     H264RawNALUnitHeader nal_unit_header;
 183
 184     uint8_t pic_parameter_set_id;
 185     uint8_t seq_parameter_set_id;
 186
 187     uint8_t entropy_coding_mode_flag;
 188     uint8_t bottom_field_pic_order_in_frame_present_flag;
 189
 190     uint8_t num_slice_groups_minus1;
 191     uint8_t slice_group_map_type;
 192     uint16_t run_length_minus1[H264_MAX_SLICE_GROUPS];
 193     uint16_t top_left[H264_MAX_SLICE_GROUPS];
 194     uint16_t bottom_right[H264_MAX_SLICE_GROUPS];
 195     uint8_t slice_group_change_direction_flag;
 196     uint16_t slice_group_change_rate_minus1;
 197     uint16_t pic_size_in_map_units_minus1;
 198
 199     uint8_t *slice_group_id;
 200     AVBufferRef *slice_group_id_ref;
 201
 202     uint8_t num_ref_idx_l0_default_active_minus1;
 203     uint8_t num_ref_idx_l1_default_active_minus1;
 204
 205     uint8_t weighted_pred_flag;
 206     uint8_t weighted_bipred_idc;
 207
 208     int8_t pic_init_qp_minus26;
 209     int8_t pic_init_qs_minus26;
 210     int8_t chroma_qp_index_offset;
 211
 212     uint8_t deblocking_filter_control_present_flag;
 213     uint8_t constrained_intra_pred_flag;
 214
 215     uint8_t more_rbsp_data;
 216
 217     uint8_t redundant_pic_cnt_present_flag;
 218     uint8_t transform_8x8_mode_flag;
 219
 220     uint8_t pic_scaling_matrix_present_flag;
 221     uint8_t pic_scaling_list_present_flag[12];
 222     H264RawScalingList scaling_list_4x4[6];
 223     H264RawScalingList scaling_list_8x8[6];
 224
 225     int8_t second_chroma_qp_index_offset;
 226 } H264RawPPS;
 227
 228 typedef struct H264RawAUD {
 229     H264RawNALUnitHeader nal_unit_header;
 230
 231     uint8_t primary_pic_type;
 232 } H264RawAUD;
 233
 234 typedef struct H264RawSEIBufferingPeriod {
 235     uint8_t seq_parameter_set_id;
 236     struct {
 237         uint32_t initial_cpb_removal_delay[H264_MAX_CPB_CNT];
 238         uint32_t initial_cpb_removal_delay_offset[H264_MAX_CPB_CNT];
 239     } nal, vcl;
 240 } H264RawSEIBufferingPeriod;
 241
 242 typedef struct H264RawSEIPicTimestamp {
 243     uint8_t ct_type;
 244     uint8_t nuit_field_based_flag;
 245     uint8_t counting_type;
 246     uint8_t full_timestamp_flag;
 247     uint8_t discontinuity_flag;
 248     uint8_t cnt_dropped_flag;
 249     uint8_t n_frames;
 250     uint8_t seconds_flag;
 251     uint8_t seconds_value;
 252     uint8_t minutes_flag;
 253     uint8_t minutes_value;
 254     uint8_t hours_flag;
 255     uint8_t hours_value;
 256     uint32_t time_offset;
 257 } H264RawSEIPicTimestamp;
 258
 259 typedef struct H264RawSEIPicTiming {
 260     uint32_t cpb_removal_delay;
 261     uint32_t dpb_output_delay;
 262     uint8_t pic_struct;
 263     uint8_t clock_timestamp_flag[3];
 264     H264RawSEIPicTimestamp timestamp[3];
 265 } H264RawSEIPicTiming;
 266
 267 typedef struct H264RawSEIPanScanRect {
 268     uint32_t pan_scan_rect_id;
 269     uint8_t  pan_scan_rect_cancel_flag;
 270     uint8_t  pan_scan_cnt_minus1;
 271     int32_t  pan_scan_rect_left_offset[3];
 272     int32_t  pan_scan_rect_right_offset[3];
 273     int32_t  pan_scan_rect_top_offset[3];
 274     int32_t  pan_scan_rect_bottom_offset[3];
 275     uint16_t pan_scan_rect_repetition_period;
 276 } H264RawSEIPanScanRect;
 277
 278 typedef struct H264RawSEIUserDataRegistered {
 279     uint8_t itu_t_t35_country_code;
 280     uint8_t itu_t_t35_country_code_extension_byte;
 281     uint8_t *data;
 282     size_t data_length;
 283     AVBufferRef *data_ref;
 284 } H264RawSEIUserDataRegistered;
 285
 286 typedef struct H264RawSEIUserDataUnregistered {
 287     uint8_t uuid_iso_iec_11578[16];
 288     uint8_t *data;
 289     size_t data_length;
 290     AVBufferRef *data_ref;
 291 } H264RawSEIUserDataUnregistered;
 292
 293 typedef struct H264RawSEIRecoveryPoint {
 294     uint16_t recovery_frame_cnt;
 295     uint8_t exact_match_flag;
 296     uint8_t broken_link_flag;
 297     uint8_t changing_slice_group_idc;
 298 } H264RawSEIRecoveryPoint;
 299
 300 typedef struct H264RawSEIDisplayOrientation {
 301     uint8_t display_orientation_cancel_flag;
 302     uint8_t hor_flip;
 303     uint8_t ver_flip;
 304     uint16_t anticlockwise_rotation;
 305     uint16_t display_orientation_repetition_period;
 306     uint8_t display_orientation_extension_flag;
 307 } H264RawSEIDisplayOrientation;
 308
 309 typedef struct H264RawSEIPayload {
 310     uint32_t payload_type;
 311     uint32_t payload_size;
 312     union {
 313         H264RawSEIBufferingPeriod buffering_period;
 314         H264RawSEIPicTiming pic_timing;
 315         H264RawSEIPanScanRect pan_scan_rect;
 316         // H264RawSEIFiller filler -> no fields.
 317         H264RawSEIUserDataRegistered user_data_registered;
 318         H264RawSEIUserDataUnregistered user_data_unregistered;
 319         H264RawSEIRecoveryPoint recovery_point;
 320         H264RawSEIDisplayOrientation display_orientation;
 321         struct {
 322             uint8_t *data;
 323             size_t data_length;
 324             AVBufferRef *data_ref;
 325         } other;
 326     } payload;
 327 } H264RawSEIPayload;
 328
 329 typedef struct H264RawSEI {
 330     H264RawNALUnitHeader nal_unit_header;
 331
 332     H264RawSEIPayload payload[H264_MAX_SEI_PAYLOADS];
 333     uint8_t payload_count;
 334 } H264RawSEI;
 335
 336 typedef struct H264RawSliceHeader {
 337     H264RawNALUnitHeader nal_unit_header;
 338
 339     uint32_t first_mb_in_slice;
 340     uint8_t slice_type;
 341
 342     uint8_t pic_parameter_set_id;
 343
 344     uint8_t colour_plane_id;
 345
 346     uint16_t frame_num;
 347     uint8_t field_pic_flag;
 348     uint8_t bottom_field_flag;
 349
 350     uint16_t idr_pic_id;
 351
 352     uint16_t pic_order_cnt_lsb;
 353     int32_t delta_pic_order_cnt_bottom;
 354     int32_t delta_pic_order_cnt[2];
 355
 356     uint8_t redundant_pic_cnt;
 357     uint8_t direct_spatial_mv_pred_flag;
 358
 359     uint8_t num_ref_idx_active_override_flag;
 360     uint8_t num_ref_idx_l0_active_minus1;
 361     uint8_t num_ref_idx_l1_active_minus1;
 362
 363     uint8_t ref_pic_list_modification_flag_l0;
 364     uint8_t ref_pic_list_modification_flag_l1;
 365     struct {
 366         uint8_t modification_of_pic_nums_idc;
 367         int32_t abs_diff_pic_num_minus1;
 368         uint8_t long_term_pic_num;
 369     } rplm_l0[H264_MAX_RPLM_COUNT], rplm_l1[H264_MAX_RPLM_COUNT];
 370
 371     uint8_t luma_log2_weight_denom;
 372     uint8_t chroma_log2_weight_denom;
 373
 374     uint8_t luma_weight_l0_flag[H264_MAX_REFS];
 375     int8_t luma_weight_l0[H264_MAX_REFS];
 376     int8_t luma_offset_l0[H264_MAX_REFS];
 377     uint8_t chroma_weight_l0_flag[H264_MAX_REFS];
 378     int8_t chroma_weight_l0[H264_MAX_REFS][2];
 379     int8_t chroma_offset_l0[H264_MAX_REFS][2];
 380
 381     uint8_t luma_weight_l1_flag[H264_MAX_REFS];
 382     int8_t luma_weight_l1[H264_MAX_REFS];
 383     int8_t luma_offset_l1[H264_MAX_REFS];
 384     uint8_t chroma_weight_l1_flag[H264_MAX_REFS];
 385     int8_t chroma_weight_l1[H264_MAX_REFS][2];
 386     int8_t chroma_offset_l1[H264_MAX_REFS][2];
 387
 388     uint8_t no_output_of_prior_pics_flag;
 389     uint8_t long_term_reference_flag;
 390
 391     uint8_t adaptive_ref_pic_marking_mode_flag;
 392     struct {
 393         uint8_t memory_management_control_operation;
 394         int32_t difference_of_pic_nums_minus1;
 395         uint8_t long_term_pic_num;
 396         uint8_t long_term_frame_idx;
 397         uint8_t max_long_term_frame_idx_plus1;
 398     } mmco[H264_MAX_MMCO_COUNT];
 399
 400     uint8_t cabac_init_idc;
 401
 402     int8_t slice_qp_delta;
 403
 404     uint8_t sp_for_switch_flag;
 405     int8_t slice_qs_delta;
 406
 407     uint8_t disable_deblocking_filter_idc;
 408     int8_t slice_alpha_c0_offset_div2;
 409     int8_t slice_beta_offset_div2;
 410
 411     uint16_t slice_group_change_cycle;
 412 } H264RawSliceHeader;
 413
 414 typedef struct H264RawSlice {
 415     H264RawSliceHeader header;
 416
 417     uint8_t *data;
 418     size_t   data_size;
 419     int      data_bit_start;
 420     AVBufferRef *data_ref;
 421 } H264RawSlice;
 422
 423 typedef struct H264RawFiller {
 424     H264RawNALUnitHeader nal_unit_header;
 425
 426     uint32_t filler_size;
 427 } H264RawFiller;
 428
 429
 430 typedef struct CodedBitstreamH264Context {
 431     // Reader/writer context in common with the H.265 implementation.
 432     CodedBitstreamH2645Context common;
 433
 434     // All currently available parameter sets.  These are updated when
 435     // any parameter set NAL unit is read/written with this context.
 436     AVBufferRef *sps_ref[H264_MAX_SPS_COUNT];
 437     AVBufferRef *pps_ref[H264_MAX_PPS_COUNT];
 438     H264RawSPS *sps[H264_MAX_SPS_COUNT];
 439     H264RawPPS *pps[H264_MAX_PPS_COUNT];
 440
 441     // The currently active parameter sets.  These are updated when any
 442     // NAL unit refers to the relevant parameter set.  These pointers
 443     // must also be present in the arrays above.
 444     const H264RawSPS *active_sps;
 445     const H264RawPPS *active_pps;
 446
 447     // The NAL unit type of the most recent normal slice.  This is required
 448     // to be able to read/write auxiliary slices, because IdrPicFlag is
 449     // otherwise unknown.
 450     uint8_t last_slice_nal_unit_type;
 451 } CodedBitstreamH264Context;
 452
 453
 454 /**
 455  * Add an SEI message to an access unit.
 456  */
 457 int ff_cbs_h264_add_sei_message(CodedBitstreamContext *ctx,
 458                                 CodedBitstreamFragment *access_unit,
 459                                 const H264RawSEIPayload *payload);
 460
 461 /**
 462  * Delete an SEI message from an access unit.
 463  *
 464  * Deletes from nal_unit, which must be an SEI NAL unit.  If this is the
 465  * last message in nal_unit, also deletes it from access_unit.
 466  */
 467 int ff_cbs_h264_delete_sei_message(CodedBitstreamContext *ctx,
 468                                    CodedBitstreamFragment *access_unit,
 469                                    CodedBitstreamUnit *nal_unit,
 470                                    int position);
 471
 472 #endif /* AVCODEC_CBS_H264_H */