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 nal_ref_idc;
  42     uint8_t nal_unit_type;
  43
  44     uint8_t svc_extension_flag;
  45     uint8_t avc_3d_extension_flag;
  46 } H264RawNALUnitHeader;
  47
  48 typedef struct H264RawScalingList {
  49     int8_t delta_scale[64];
  50 } H264RawScalingList;
  51
  52 typedef struct H264RawHRD {
  53     uint8_t cpb_cnt_minus1;
  54     uint8_t bit_rate_scale;
  55     uint8_t cpb_size_scale;
  56
  57     uint32_t bit_rate_value_minus1[H264_MAX_CPB_CNT];
  58     uint32_t cpb_size_value_minus1[H264_MAX_CPB_CNT];
  59     uint8_t cbr_flag[H264_MAX_CPB_CNT];
  60
  61     uint8_t initial_cpb_removal_delay_length_minus1;
  62     uint8_t cpb_removal_delay_length_minus1;
  63     uint8_t dpb_output_delay_length_minus1;
  64     uint8_t time_offset_length;
  65 } H264RawHRD;
  66
  67 typedef struct H264RawVUI {
  68     uint8_t aspect_ratio_info_present_flag;
  69     uint8_t aspect_ratio_idc;
  70     uint16_t sar_width;
  71     uint16_t sar_height;
  72
  73     uint8_t overscan_info_present_flag;
  74     uint8_t overscan_appropriate_flag;
  75
  76     uint8_t video_signal_type_present_flag;
  77     uint8_t video_format;
  78     uint8_t video_full_range_flag;
  79     uint8_t colour_description_present_flag;
  80     uint8_t colour_primaries;
  81     uint8_t transfer_characteristics;
  82     uint8_t matrix_coefficients;
  83
  84     uint8_t chroma_loc_info_present_flag;
  85     uint8_t chroma_sample_loc_type_top_field;
  86     uint8_t chroma_sample_loc_type_bottom_field;
  87
  88     uint8_t timing_info_present_flag;
  89     uint32_t num_units_in_tick;
  90     uint32_t time_scale;
  91     uint8_t fixed_frame_rate_flag;
  92
  93     uint8_t nal_hrd_parameters_present_flag;
  94     H264RawHRD nal_hrd_parameters;
  95     uint8_t vcl_hrd_parameters_present_flag;
  96     H264RawHRD vcl_hrd_parameters;
  97     uint8_t low_delay_hrd_flag;
  98
  99     uint8_t pic_struct_present_flag;
 100
 101     uint8_t bitstream_restriction_flag;
 102     uint8_t motion_vectors_over_pic_boundaries_flag;
 103     uint8_t max_bytes_per_pic_denom;
 104     uint8_t max_bits_per_mb_denom;
 105     uint8_t log2_max_mv_length_horizontal;
 106     uint8_t log2_max_mv_length_vertical;
 107     uint8_t max_num_reorder_frames;
 108     uint8_t max_dec_frame_buffering;
 109 } H264RawVUI;
 110
 111 typedef struct H264RawSPS {
 112     H264RawNALUnitHeader nal_unit_header;
 113
 114     uint8_t profile_idc;
 115     uint8_t constraint_set0_flag;
 116     uint8_t constraint_set1_flag;
 117     uint8_t constraint_set2_flag;
 118     uint8_t constraint_set3_flag;
 119     uint8_t constraint_set4_flag;
 120     uint8_t constraint_set5_flag;
 121     uint8_t reserved_zero_2bits;
 122     uint8_t level_idc;
 123
 124     uint8_t seq_parameter_set_id;
 125
 126     uint8_t chroma_format_idc;
 127     uint8_t separate_colour_plane_flag;
 128     uint8_t bit_depth_luma_minus8;
 129     uint8_t bit_depth_chroma_minus8;
 130     uint8_t qpprime_y_zero_transform_bypass_flag;
 131
 132     uint8_t seq_scaling_matrix_present_flag;
 133     uint8_t seq_scaling_list_present_flag[12];
 134     H264RawScalingList scaling_list_4x4[6];
 135     H264RawScalingList scaling_list_8x8[6];
 136
 137     uint8_t log2_max_frame_num_minus4;
 138     uint8_t pic_order_cnt_type;
 139     uint8_t log2_max_pic_order_cnt_lsb_minus4;
 140     uint8_t delta_pic_order_always_zero_flag;
 141     int32_t offset_for_non_ref_pic;
 142     int32_t offset_for_top_to_bottom_field;
 143     uint8_t num_ref_frames_in_pic_order_cnt_cycle;
 144     int32_t offset_for_ref_frame[256];
 145
 146     uint8_t max_num_ref_frames;
 147     uint8_t gaps_in_frame_num_allowed_flag;
 148
 149     uint16_t pic_width_in_mbs_minus1;
 150     uint16_t pic_height_in_map_units_minus1;
 151
 152     uint8_t frame_mbs_only_flag;
 153     uint8_t mb_adaptive_frame_field_flag;
 154     uint8_t direct_8x8_inference_flag;
 155
 156     uint8_t frame_cropping_flag;
 157     uint16_t frame_crop_left_offset;
 158     uint16_t frame_crop_right_offset;
 159     uint16_t frame_crop_top_offset;
 160     uint16_t frame_crop_bottom_offset;
 161
 162     uint8_t vui_parameters_present_flag;
 163     H264RawVUI vui;
 164 } H264RawSPS;
 165
 166 typedef struct H264RawSPSExtension {
 167     H264RawNALUnitHeader nal_unit_header;
 168
 169     uint8_t seq_parameter_set_id;
 170
 171     uint8_t aux_format_idc;
 172     uint8_t bit_depth_aux_minus8;
 173     uint8_t alpha_incr_flag;
 174     uint16_t alpha_opaque_value;
 175     uint16_t alpha_transparent_value;
 176
 177     uint8_t additional_extension_flag;
 178 } H264RawSPSExtension;
 179
 180 typedef struct H264RawPPS {
 181     H264RawNALUnitHeader nal_unit_header;
 182
 183     uint8_t pic_parameter_set_id;
 184     uint8_t seq_parameter_set_id;
 185
 186     uint8_t entropy_coding_mode_flag;
 187     uint8_t bottom_field_pic_order_in_frame_present_flag;
 188
 189     uint8_t num_slice_groups_minus1;
 190     uint8_t slice_group_map_type;
 191     uint16_t run_length_minus1[H264_MAX_SLICE_GROUPS];
 192     uint16_t top_left[H264_MAX_SLICE_GROUPS];
 193     uint16_t bottom_right[H264_MAX_SLICE_GROUPS];
 194     uint8_t slice_group_change_direction_flag;
 195     uint16_t slice_group_change_rate_minus1;
 196     uint16_t pic_size_in_map_units_minus1;
 197
 198     uint8_t *slice_group_id;
 199     AVBufferRef *slice_group_id_ref;
 200
 201     uint8_t num_ref_idx_l0_default_active_minus1;
 202     uint8_t num_ref_idx_l1_default_active_minus1;
 203
 204     uint8_t weighted_pred_flag;
 205     uint8_t weighted_bipred_idc;
 206
 207     int8_t pic_init_qp_minus26;
 208     int8_t pic_init_qs_minus26;
 209     int8_t chroma_qp_index_offset;
 210
 211     uint8_t deblocking_filter_control_present_flag;
 212     uint8_t constrained_intra_pred_flag;
 213
 214     uint8_t more_rbsp_data;
 215
 216     uint8_t redundant_pic_cnt_present_flag;
 217     uint8_t transform_8x8_mode_flag;
 218
 219     uint8_t pic_scaling_matrix_present_flag;
 220     uint8_t pic_scaling_list_present_flag[12];
 221     H264RawScalingList scaling_list_4x4[6];
 222     H264RawScalingList scaling_list_8x8[6];
 223
 224     int8_t second_chroma_qp_index_offset;
 225 } H264RawPPS;
 226
 227 typedef struct H264RawAUD {
 228     H264RawNALUnitHeader nal_unit_header;
 229
 230     uint8_t primary_pic_type;
 231 } H264RawAUD;
 232
 233 typedef struct H264RawSEIBufferingPeriod {
 234     uint8_t seq_parameter_set_id;
 235     struct {
 236         uint32_t initial_cpb_removal_delay[H264_MAX_CPB_CNT];
 237         uint32_t initial_cpb_removal_delay_offset[H264_MAX_CPB_CNT];
 238     } nal, vcl;
 239 } H264RawSEIBufferingPeriod;
 240
 241 typedef struct H264RawSEIPicTimestamp {
 242     uint8_t ct_type;
 243     uint8_t nuit_field_based_flag;
 244     uint8_t counting_type;
 245     uint8_t full_timestamp_flag;
 246     uint8_t discontinuity_flag;
 247     uint8_t cnt_dropped_flag;
 248     uint8_t n_frames;
 249     uint8_t seconds_flag;
 250     uint8_t seconds_value;
 251     uint8_t minutes_flag;
 252     uint8_t minutes_value;
 253     uint8_t hours_flag;
 254     uint8_t hours_value;
 255     int32_t time_offset;
 256 } H264RawSEIPicTimestamp;
 257
 258 typedef struct H264RawSEIPicTiming {
 259     uint32_t cpb_removal_delay;
 260     uint32_t dpb_output_delay;
 261     uint8_t pic_struct;
 262     uint8_t clock_timestamp_flag[3];
 263     H264RawSEIPicTimestamp timestamp[3];
 264 } H264RawSEIPicTiming;
 265
 266 typedef struct H264RawSEIPanScanRect {
 267     uint32_t pan_scan_rect_id;
 268     uint8_t  pan_scan_rect_cancel_flag;
 269     uint8_t  pan_scan_cnt_minus1;
 270     int32_t  pan_scan_rect_left_offset[3];
 271     int32_t  pan_scan_rect_right_offset[3];
 272     int32_t  pan_scan_rect_top_offset[3];
 273     int32_t  pan_scan_rect_bottom_offset[3];
 274     uint16_t pan_scan_rect_repetition_period;
 275 } H264RawSEIPanScanRect;
 276
 277 typedef struct H264RawSEIUserDataRegistered {
 278     uint8_t itu_t_t35_country_code;
 279     uint8_t itu_t_t35_country_code_extension_byte;
 280     uint8_t     *data;
 281     AVBufferRef *data_ref;
 282     size_t       data_length;
 283 } H264RawSEIUserDataRegistered;
 284
 285 typedef struct H264RawSEIUserDataUnregistered {
 286     uint8_t uuid_iso_iec_11578[16];
 287     uint8_t     *data;
 288     AVBufferRef *data_ref;
 289     size_t       data_length;
 290 } H264RawSEIUserDataUnregistered;
 291
 292 typedef struct H264RawSEIRecoveryPoint {
 293     uint16_t recovery_frame_cnt;
 294     uint8_t exact_match_flag;
 295     uint8_t broken_link_flag;
 296     uint8_t changing_slice_group_idc;
 297 } H264RawSEIRecoveryPoint;
 298
 299 typedef struct H264RawSEIDisplayOrientation {
 300     uint8_t display_orientation_cancel_flag;
 301     uint8_t hor_flip;
 302     uint8_t ver_flip;
 303     uint16_t anticlockwise_rotation;
 304     uint16_t display_orientation_repetition_period;
 305     uint8_t display_orientation_extension_flag;
 306 } H264RawSEIDisplayOrientation;
 307
 308 typedef struct H264RawSEIMasteringDisplayColourVolume {
 309     uint16_t display_primaries_x[3];
 310     uint16_t display_primaries_y[3];
 311     uint16_t white_point_x;
 312     uint16_t white_point_y;
 313     uint32_t max_display_mastering_luminance;
 314     uint32_t min_display_mastering_luminance;
 315 } H264RawSEIMasteringDisplayColourVolume;
 316
 317 typedef struct H264RawSEIAlternativeTransferCharacteristics {
 318     uint8_t preferred_transfer_characteristics;
 319 } H264RawSEIAlternativeTransferCharacteristics;
 320
 321 typedef struct H264RawSEIPayload {
 322     uint32_t payload_type;
 323     uint32_t payload_size;
 324     union {
 325         H264RawSEIBufferingPeriod buffering_period;
 326         H264RawSEIPicTiming pic_timing;
 327         H264RawSEIPanScanRect pan_scan_rect;
 328         // H264RawSEIFiller filler -> no fields.
 329         H264RawSEIUserDataRegistered user_data_registered;
 330         H264RawSEIUserDataUnregistered user_data_unregistered;
 331         H264RawSEIRecoveryPoint recovery_point;
 332         H264RawSEIDisplayOrientation display_orientation;
 333         H264RawSEIMasteringDisplayColourVolume mastering_display_colour_volume;
 334         H264RawSEIAlternativeTransferCharacteristics
 335             alternative_transfer_characteristics;
 336         struct {
 337             uint8_t     *data;
 338             AVBufferRef *data_ref;
 339             size_t       data_length;
 340         } other;
 341     } payload;
 342 } H264RawSEIPayload;
 343
 344 typedef struct H264RawSEI {
 345     H264RawNALUnitHeader nal_unit_header;
 346
 347     H264RawSEIPayload payload[H264_MAX_SEI_PAYLOADS];
 348     uint8_t payload_count;
 349 } H264RawSEI;
 350
 351 typedef struct H264RawSliceHeader {
 352     H264RawNALUnitHeader nal_unit_header;
 353
 354     uint32_t first_mb_in_slice;
 355     uint8_t slice_type;
 356
 357     uint8_t pic_parameter_set_id;
 358
 359     uint8_t colour_plane_id;
 360
 361     uint16_t frame_num;
 362     uint8_t field_pic_flag;
 363     uint8_t bottom_field_flag;
 364
 365     uint16_t idr_pic_id;
 366
 367     uint16_t pic_order_cnt_lsb;
 368     int32_t delta_pic_order_cnt_bottom;
 369     int32_t delta_pic_order_cnt[2];
 370
 371     uint8_t redundant_pic_cnt;
 372     uint8_t direct_spatial_mv_pred_flag;
 373
 374     uint8_t num_ref_idx_active_override_flag;
 375     uint8_t num_ref_idx_l0_active_minus1;
 376     uint8_t num_ref_idx_l1_active_minus1;
 377
 378     uint8_t ref_pic_list_modification_flag_l0;
 379     uint8_t ref_pic_list_modification_flag_l1;
 380     struct {
 381         uint8_t modification_of_pic_nums_idc;
 382         int32_t abs_diff_pic_num_minus1;
 383         uint8_t long_term_pic_num;
 384     } rplm_l0[H264_MAX_RPLM_COUNT], rplm_l1[H264_MAX_RPLM_COUNT];
 385
 386     uint8_t luma_log2_weight_denom;
 387     uint8_t chroma_log2_weight_denom;
 388
 389     uint8_t luma_weight_l0_flag[H264_MAX_REFS];
 390     int8_t luma_weight_l0[H264_MAX_REFS];
 391     int8_t luma_offset_l0[H264_MAX_REFS];
 392     uint8_t chroma_weight_l0_flag[H264_MAX_REFS];
 393     int8_t chroma_weight_l0[H264_MAX_REFS][2];
 394     int8_t chroma_offset_l0[H264_MAX_REFS][2];
 395
 396     uint8_t luma_weight_l1_flag[H264_MAX_REFS];
 397     int8_t luma_weight_l1[H264_MAX_REFS];
 398     int8_t luma_offset_l1[H264_MAX_REFS];
 399     uint8_t chroma_weight_l1_flag[H264_MAX_REFS];
 400     int8_t chroma_weight_l1[H264_MAX_REFS][2];
 401     int8_t chroma_offset_l1[H264_MAX_REFS][2];
 402
 403     uint8_t no_output_of_prior_pics_flag;
 404     uint8_t long_term_reference_flag;
 405
 406     uint8_t adaptive_ref_pic_marking_mode_flag;
 407     struct {
 408         uint8_t memory_management_control_operation;
 409         int32_t difference_of_pic_nums_minus1;
 410         uint8_t long_term_pic_num;
 411         uint8_t long_term_frame_idx;
 412         uint8_t max_long_term_frame_idx_plus1;
 413     } mmco[H264_MAX_MMCO_COUNT];
 414
 415     uint8_t cabac_init_idc;
 416
 417     int8_t slice_qp_delta;
 418
 419     uint8_t sp_for_switch_flag;
 420     int8_t slice_qs_delta;
 421
 422     uint8_t disable_deblocking_filter_idc;
 423     int8_t slice_alpha_c0_offset_div2;
 424     int8_t slice_beta_offset_div2;
 425
 426     uint16_t slice_group_change_cycle;
 427 } H264RawSliceHeader;
 428
 429 typedef struct H264RawSlice {
 430     H264RawSliceHeader header;
 431
 432     uint8_t     *data;
 433     AVBufferRef *data_ref;
 434     size_t       data_size;
 435     int          data_bit_start;
 436 } H264RawSlice;
 437
 438 typedef struct H264RawFiller {
 439     H264RawNALUnitHeader nal_unit_header;
 440
 441     uint32_t filler_size;
 442 } H264RawFiller;
 443
 444
 445 typedef struct CodedBitstreamH264Context {
 446     // Reader/writer context in common with the H.265 implementation.
 447     CodedBitstreamH2645Context common;
 448
 449     // All currently available parameter sets.  These are updated when
 450     // any parameter set NAL unit is read/written with this context.
 451     AVBufferRef *sps_ref[H264_MAX_SPS_COUNT];
 452     AVBufferRef *pps_ref[H264_MAX_PPS_COUNT];
 453     H264RawSPS *sps[H264_MAX_SPS_COUNT];
 454     H264RawPPS *pps[H264_MAX_PPS_COUNT];
 455
 456     // The currently active parameter sets.  These are updated when any
 457     // NAL unit refers to the relevant parameter set.  These pointers
 458     // must also be present in the arrays above.
 459     const H264RawSPS *active_sps;
 460     const H264RawPPS *active_pps;
 461
 462     // The NAL unit type of the most recent normal slice.  This is required
 463     // to be able to read/write auxiliary slices, because IdrPicFlag is
 464     // otherwise unknown.
 465     uint8_t last_slice_nal_unit_type;
 466 } CodedBitstreamH264Context;
 467
 468
 469 /**
 470  * Add an SEI message to an access unit.
 471  *
 472  * On success, the payload will be owned by a unit in access_unit;
 473  * on failure, the content of the payload will be freed.
 474  */
 475 int ff_cbs_h264_add_sei_message(CodedBitstreamFragment *access_unit,
 476                                 H264RawSEIPayload *payload);
 477
 478 /**
 479  * Delete an SEI message from an access unit.
 480  *
 481  * Deletes from nal_unit, which must be an SEI NAL unit.  If this is the
 482  * last message in nal_unit, also deletes it from access_unit.
 483  *
 484  * Requires nal_unit to be a unit in access_unit and position to be >= 0
 485  * and < the payload count of the SEI nal_unit.
 486  */
 487 void ff_cbs_h264_delete_sei_message(CodedBitstreamFragment *access_unit,
 488                                     CodedBitstreamUnit *nal_unit,
 489                                     int position);
 490
 491 #endif /* AVCODEC_CBS_H264_H */