X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fhevc.h;h=bb4f6605866a2c23cb1f837ce5d3fcc30fc93fd6;hb=fa1749dd34c55fb997c97dfc4da9383c9976ab91;hp=ab95035ddadbb5783fbd2cdae65f203ed4e03b1a;hpb=064698d381e1e7790f21b0199a8930ea04e2e942;p=ffmpeg diff --git a/libavcodec/hevc.h b/libavcodec/hevc.h index ab95035ddad..bb4f6605866 100644 --- a/libavcodec/hevc.h +++ b/libavcodec/hevc.h @@ -1,7 +1,5 @@ /* - * HEVC video decoder - * - * Copyright (C) 2012 - 2013 Guillaume Martres + * HEVC shared code * * This file is part of Libav. * @@ -23,1044 +21,56 @@ #ifndef AVCODEC_HEVC_H #define AVCODEC_HEVC_H -#include -#include - -#include "libavutil/buffer.h" -#include "libavutil/md5.h" - -#include "avcodec.h" -#include "cabac.h" -#include "dsputil.h" -#include "get_bits.h" -#include "internal.h" -#include "thread.h" -#include "videodsp.h" - -#define MAX_DPB_SIZE 16 // A.4.1 -#define MAX_REFS 16 - -/** - * 7.4.2.1 - */ -#define MAX_SUB_LAYERS 7 -#define MAX_VPS_COUNT 16 -#define MAX_SPS_COUNT 32 -#define MAX_PPS_COUNT 256 -#define MAX_SHORT_TERM_RPS_COUNT 64 -#define MAX_CU_SIZE 128 - -//TODO: check if this is really the maximum -#define MAX_TRANSFORM_DEPTH 5 - -#define MAX_TB_SIZE 32 -#define MAX_PB_SIZE 64 -#define MAX_LOG2_CTB_SIZE 6 -#define MAX_QP 51 -#define DEFAULT_INTRA_TC_OFFSET 2 - -#define HEVC_CONTEXTS 183 - -#define MRG_MAX_NUM_CANDS 5 - -#define L0 0 -#define L1 1 - -#define EPEL_EXTRA_BEFORE 1 -#define EPEL_EXTRA_AFTER 2 -#define EPEL_EXTRA 3 - -/** - * Value of the luma sample at position (x, y) in the 2D array tab. - */ -#define SAMPLE(tab, x, y) ((tab)[(y) * s->sps->width + (x)]) -#define SAMPLE_CTB(tab, x, y) ((tab)[(y) * min_cb_width + (x)]) -#define SAMPLE_CBF(tab, x, y) ((tab)[((y) & ((1<nal_unit_type == NAL_IDR_W_RADL || s->nal_unit_type == NAL_IDR_N_LP) -#define IS_BLA(s) (s->nal_unit_type == NAL_BLA_W_RADL || s->nal_unit_type == NAL_BLA_W_LP || \ - s->nal_unit_type == NAL_BLA_N_LP) -#define IS_IRAP(s) (s->nal_unit_type >= 16 && s->nal_unit_type <= 23) - /** * Table 7-3: NAL unit type codes */ -enum NALUnitType { - NAL_TRAIL_N = 0, - NAL_TRAIL_R = 1, - NAL_TSA_N = 2, - NAL_TSA_R = 3, - NAL_STSA_N = 4, - NAL_STSA_R = 5, - NAL_RADL_N = 6, - NAL_RADL_R = 7, - NAL_RASL_N = 8, - NAL_RASL_R = 9, - NAL_BLA_W_LP = 16, - NAL_BLA_W_RADL = 17, - NAL_BLA_N_LP = 18, - NAL_IDR_W_RADL = 19, - NAL_IDR_N_LP = 20, - NAL_CRA_NUT = 21, - NAL_VPS = 32, - NAL_SPS = 33, - NAL_PPS = 34, - NAL_AUD = 35, - NAL_EOS_NUT = 36, - NAL_EOB_NUT = 37, - NAL_FD_NUT = 38, - NAL_SEI_PREFIX = 39, - NAL_SEI_SUFFIX = 40, +enum HEVCNALUnitType { + HEVC_NAL_TRAIL_N = 0, + HEVC_NAL_TRAIL_R = 1, + HEVC_NAL_TSA_N = 2, + HEVC_NAL_TSA_R = 3, + HEVC_NAL_STSA_N = 4, + HEVC_NAL_STSA_R = 5, + HEVC_NAL_RADL_N = 6, + HEVC_NAL_RADL_R = 7, + HEVC_NAL_RASL_N = 8, + HEVC_NAL_RASL_R = 9, + HEVC_NAL_BLA_W_LP = 16, + HEVC_NAL_BLA_W_RADL = 17, + HEVC_NAL_BLA_N_LP = 18, + HEVC_NAL_IDR_W_RADL = 19, + HEVC_NAL_IDR_N_LP = 20, + HEVC_NAL_CRA_NUT = 21, + HEVC_NAL_VPS = 32, + HEVC_NAL_SPS = 33, + HEVC_NAL_PPS = 34, + HEVC_NAL_AUD = 35, + HEVC_NAL_EOS_NUT = 36, + HEVC_NAL_EOB_NUT = 37, + HEVC_NAL_FD_NUT = 38, + HEVC_NAL_SEI_PREFIX = 39, + HEVC_NAL_SEI_SUFFIX = 40, }; -enum RPSType { - ST_CURR_BEF = 0, - ST_CURR_AFT, - ST_FOLL, - LT_CURR, - LT_FOLL, - NB_RPS_TYPE, +enum HEVCSliceType { + HEVC_SLICE_B = 0, + HEVC_SLICE_P = 1, + HEVC_SLICE_I = 2, }; -enum SliceType { - B_SLICE = 0, - P_SLICE = 1, - I_SLICE = 2, -}; - -enum SyntaxElement { - SAO_MERGE_FLAG = 0, - SAO_TYPE_IDX, - SAO_EO_CLASS, - SAO_BAND_POSITION, - SAO_OFFSET_ABS, - SAO_OFFSET_SIGN, - END_OF_SLICE_FLAG, - SPLIT_CODING_UNIT_FLAG, - CU_TRANSQUANT_BYPASS_FLAG, - SKIP_FLAG, - CU_QP_DELTA, - PRED_MODE_FLAG, - PART_MODE, - PCM_FLAG, - PREV_INTRA_LUMA_PRED_FLAG, - MPM_IDX, - REM_INTRA_LUMA_PRED_MODE, - INTRA_CHROMA_PRED_MODE, - MERGE_FLAG, - MERGE_IDX, - INTER_PRED_IDC, - REF_IDX_L0, - REF_IDX_L1, - ABS_MVD_GREATER0_FLAG, - ABS_MVD_GREATER1_FLAG, - ABS_MVD_MINUS2, - MVD_SIGN_FLAG, - MVP_LX_FLAG, - NO_RESIDUAL_DATA_FLAG, - SPLIT_TRANSFORM_FLAG, - CBF_LUMA, - CBF_CB_CR, - TRANSFORM_SKIP_FLAG, - LAST_SIGNIFICANT_COEFF_X_PREFIX, - LAST_SIGNIFICANT_COEFF_Y_PREFIX, - LAST_SIGNIFICANT_COEFF_X_SUFFIX, - LAST_SIGNIFICANT_COEFF_Y_SUFFIX, - SIGNIFICANT_COEFF_GROUP_FLAG, - SIGNIFICANT_COEFF_FLAG, - COEFF_ABS_LEVEL_GREATER1_FLAG, - COEFF_ABS_LEVEL_GREATER2_FLAG, - COEFF_ABS_LEVEL_REMAINING, - COEFF_SIGN_FLAG, -}; - -enum PartMode { - PART_2Nx2N = 0, - PART_2NxN = 1, - PART_Nx2N = 2, - PART_NxN = 3, - PART_2NxnU = 4, - PART_2NxnD = 5, - PART_nLx2N = 6, - PART_nRx2N = 7, -}; - -enum PredMode { - MODE_INTER = 0, - MODE_INTRA, - MODE_SKIP, -}; - -enum InterPredIdc { - PRED_L0 = 0, - PRED_L1, - PRED_BI, -}; - -enum IntraPredMode { - INTRA_PLANAR = 0, - INTRA_DC, - INTRA_ANGULAR_2, - INTRA_ANGULAR_3, - INTRA_ANGULAR_4, - INTRA_ANGULAR_5, - INTRA_ANGULAR_6, - INTRA_ANGULAR_7, - INTRA_ANGULAR_8, - INTRA_ANGULAR_9, - INTRA_ANGULAR_10, - INTRA_ANGULAR_11, - INTRA_ANGULAR_12, - INTRA_ANGULAR_13, - INTRA_ANGULAR_14, - INTRA_ANGULAR_15, - INTRA_ANGULAR_16, - INTRA_ANGULAR_17, - INTRA_ANGULAR_18, - INTRA_ANGULAR_19, - INTRA_ANGULAR_20, - INTRA_ANGULAR_21, - INTRA_ANGULAR_22, - INTRA_ANGULAR_23, - INTRA_ANGULAR_24, - INTRA_ANGULAR_25, - INTRA_ANGULAR_26, - INTRA_ANGULAR_27, - INTRA_ANGULAR_28, - INTRA_ANGULAR_29, - INTRA_ANGULAR_30, - INTRA_ANGULAR_31, - INTRA_ANGULAR_32, - INTRA_ANGULAR_33, - INTRA_ANGULAR_34, -}; - -enum SAOType { - SAO_NOT_APPLIED = 0, - SAO_BAND, - SAO_EDGE, -}; - -enum SAOEOClass { - SAO_EO_HORIZ = 0, - SAO_EO_VERT, - SAO_EO_135D, - SAO_EO_45D, -}; - -enum ScanType { - SCAN_DIAG = 0, - SCAN_HORIZ, - SCAN_VERT, -}; - -typedef struct ShortTermRPS { - int num_negative_pics; - int num_delta_pocs; - int32_t delta_poc[32]; - uint8_t used[32]; -} ShortTermRPS; - -typedef struct LongTermRPS { - int poc[32]; - uint8_t used[32]; - uint8_t nb_refs; -} LongTermRPS; - -typedef struct RefPicList { - struct HEVCFrame *ref[MAX_REFS]; - int list[MAX_REFS]; - int isLongTerm[MAX_REFS]; - int nb_refs; -} RefPicList; - -typedef struct RefPicListTab { - RefPicList refPicList[2]; -} RefPicListTab; - -typedef struct HEVCWindow { - int left_offset; - int right_offset; - int top_offset; - int bottom_offset; -} HEVCWindow; - -typedef struct VUI { - AVRational sar; - - int overscan_info_present_flag; - int overscan_appropriate_flag; - - int video_signal_type_present_flag; - int video_format; - int video_full_range_flag; - int colour_description_present_flag; - uint8_t colour_primaries; - uint8_t transfer_characteristic; - uint8_t matrix_coeffs; - - int chroma_loc_info_present_flag; - int chroma_sample_loc_type_top_field; - int chroma_sample_loc_type_bottom_field; - int neutra_chroma_indication_flag; - - int field_seq_flag; - int frame_field_info_present_flag; - - int default_display_window_flag; - HEVCWindow def_disp_win; - - int vui_timing_info_present_flag; - uint32_t vui_num_units_in_tick; - uint32_t vui_time_scale; - int vui_poc_proportional_to_timing_flag; - int vui_num_ticks_poc_diff_one_minus1; - int vui_hrd_parameters_present_flag; - - int bitstream_restriction_flag; - int tiles_fixed_structure_flag; - int motion_vectors_over_pic_boundaries_flag; - int restricted_ref_pic_lists_flag; - int min_spatial_segmentation_idc; - int max_bytes_per_pic_denom; - int max_bits_per_min_cu_denom; - int log2_max_mv_length_horizontal; - int log2_max_mv_length_vertical; -} VUI; - -typedef struct PTL { - int general_profile_space; - uint8_t general_tier_flag; - int general_profile_idc; - int general_profile_compatibility_flag[32]; - int general_level_idc; - - uint8_t sub_layer_profile_present_flag[MAX_SUB_LAYERS]; - uint8_t sub_layer_level_present_flag[MAX_SUB_LAYERS]; - - int sub_layer_profile_space[MAX_SUB_LAYERS]; - uint8_t sub_layer_tier_flag[MAX_SUB_LAYERS]; - int sub_layer_profile_idc[MAX_SUB_LAYERS]; - uint8_t sub_layer_profile_compatibility_flags[MAX_SUB_LAYERS][32]; - int sub_layer_level_idc[MAX_SUB_LAYERS]; -} PTL; - -typedef struct HEVCVPS { - uint8_t vps_temporal_id_nesting_flag; - int vps_max_layers; - int vps_max_sub_layers; ///< vps_max_temporal_layers_minus1 + 1 - - PTL ptl; - int vps_sub_layer_ordering_info_present_flag; - unsigned int vps_max_dec_pic_buffering[MAX_SUB_LAYERS]; - unsigned int vps_num_reorder_pics[MAX_SUB_LAYERS]; - unsigned int vps_max_latency_increase[MAX_SUB_LAYERS]; - int vps_max_layer_id; - int vps_num_layer_sets; ///< vps_num_layer_sets_minus1 + 1 - uint8_t vps_timing_info_present_flag; - uint32_t vps_num_units_in_tick; - uint32_t vps_time_scale; - uint8_t vps_poc_proportional_to_timing_flag; - int vps_num_ticks_poc_diff_one; ///< vps_num_ticks_poc_diff_one_minus1 + 1 - int vps_num_hrd_parameters; -} HEVCVPS; - -typedef struct ScalingList { - /* This is a little wasteful, since sizeID 0 only needs 8 coeffs, - * and size ID 3 only has 2 arrays, not 6. */ - uint8_t sl[4][6][64]; - uint8_t sl_dc[2][6]; -} ScalingList; - -typedef struct HEVCSPS { - int vps_id; - int chroma_format_idc; - uint8_t separate_colour_plane_flag; - - ///< output (i.e. cropped) values - int output_width, output_height; - HEVCWindow output_window; - - HEVCWindow pic_conf_win; - - int bit_depth; - int pixel_shift; - enum AVPixelFormat pix_fmt; - - unsigned int log2_max_poc_lsb; - int pcm_enabled_flag; - - int max_sub_layers; - struct { - int max_dec_pic_buffering; - int num_reorder_pics; - int max_latency_increase; - } temporal_layer[MAX_SUB_LAYERS]; - - VUI vui; - PTL ptl; - - uint8_t scaling_list_enable_flag; - ScalingList scaling_list; - - unsigned int nb_st_rps; - ShortTermRPS st_rps[MAX_SHORT_TERM_RPS_COUNT]; - - uint8_t amp_enabled_flag; - uint8_t sao_enabled; - - uint8_t long_term_ref_pics_present_flag; - uint16_t lt_ref_pic_poc_lsb_sps[32]; - uint8_t used_by_curr_pic_lt_sps_flag[32]; - uint8_t num_long_term_ref_pics_sps; - - struct { - uint8_t bit_depth; - unsigned int log2_min_pcm_cb_size; - unsigned int log2_max_pcm_cb_size; - uint8_t loop_filter_disable_flag; - } pcm; - uint8_t sps_temporal_mvp_enabled_flag; - uint8_t sps_strong_intra_smoothing_enable_flag; - - unsigned int log2_min_cb_size; - unsigned int log2_diff_max_min_coding_block_size; - unsigned int log2_min_tb_size; - unsigned int log2_max_trafo_size; - unsigned int log2_ctb_size; - unsigned int log2_min_pu_size; - - int max_transform_hierarchy_depth_inter; - int max_transform_hierarchy_depth_intra; - - ///< coded frame dimension in various units - int width; - int height; - int ctb_width; - int ctb_height; - int ctb_size; - int min_cb_width; - int min_cb_height; - int min_tb_width; - int min_tb_height; - int min_pu_width; - int min_pu_height; - - int hshift[3]; - int vshift[3]; - - int qp_bd_offset; -} HEVCSPS; - -typedef struct HEVCPPS { - int sps_id; ///< seq_parameter_set_id - - uint8_t sign_data_hiding_flag; - - uint8_t cabac_init_present_flag; - - int num_ref_idx_l0_default_active; ///< num_ref_idx_l0_default_active_minus1 + 1 - int num_ref_idx_l1_default_active; ///< num_ref_idx_l1_default_active_minus1 + 1 - int pic_init_qp_minus26; - - uint8_t constrained_intra_pred_flag; - uint8_t transform_skip_enabled_flag; - - uint8_t cu_qp_delta_enabled_flag; - int diff_cu_qp_delta_depth; - - int cb_qp_offset; - int cr_qp_offset; - uint8_t pic_slice_level_chroma_qp_offsets_present_flag; - uint8_t weighted_pred_flag; - uint8_t weighted_bipred_flag; - uint8_t output_flag_present_flag; - uint8_t transquant_bypass_enable_flag; - - uint8_t dependent_slice_segments_enabled_flag; - uint8_t tiles_enabled_flag; - uint8_t entropy_coding_sync_enabled_flag; - - int num_tile_columns; ///< num_tile_columns_minus1 + 1 - int num_tile_rows; ///< num_tile_rows_minus1 + 1 - uint8_t uniform_spacing_flag; - uint8_t loop_filter_across_tiles_enabled_flag; - - uint8_t seq_loop_filter_across_slices_enabled_flag; - - uint8_t deblocking_filter_control_present_flag; - uint8_t deblocking_filter_override_enabled_flag; - uint8_t disable_dbf; - int beta_offset; ///< beta_offset_div2 * 2 - int tc_offset; ///< tc_offset_div2 * 2 - - uint8_t scaling_list_data_present_flag; - ScalingList scaling_list; - - uint8_t lists_modification_present_flag; - int log2_parallel_merge_level; ///< log2_parallel_merge_level_minus2 + 2 - int num_extra_slice_header_bits; - uint8_t slice_header_extension_present_flag; - - // Inferred parameters - int *column_width; ///< ColumnWidth - int *row_height; ///< RowHeight - int *col_bd; ///< ColBd - int *row_bd; ///< RowBd - int *col_idxX; - - int *ctb_addr_rs_to_ts; ///< CtbAddrRSToTS - int *ctb_addr_ts_to_rs; ///< CtbAddrTSToRS - int *tile_id; ///< TileId - int *tile_pos_rs; ///< TilePosRS - int *min_cb_addr_zs; ///< MinCbAddrZS - int *min_tb_addr_zs; ///< MinTbAddrZS -} HEVCPPS; - -typedef struct SliceHeader { - int pps_id; - - ///< address (in raster order) of the first block in the current slice segment - unsigned int slice_segment_addr; - ///< address (in raster order) of the first block in the current slice - unsigned int slice_addr; - - enum SliceType slice_type; - - int pic_order_cnt_lsb; - - uint8_t first_slice_in_pic_flag; - uint8_t dependent_slice_segment_flag; - uint8_t pic_output_flag; - uint8_t colour_plane_id; - - ///< RPS coded in the slice header itself is stored here - ShortTermRPS slice_rps; - const ShortTermRPS *short_term_rps; - LongTermRPS long_term_rps; - unsigned int list_entry_lx[2][32]; - - uint8_t rpl_modification_flag[2]; - uint8_t no_output_of_prior_pics_flag; - uint8_t slice_temporal_mvp_enabled_flag; - - unsigned int nb_refs[2]; - - uint8_t slice_sample_adaptive_offset_flag[3]; - uint8_t mvd_l1_zero_flag; - - uint8_t cabac_init_flag; - uint8_t disable_deblocking_filter_flag; ///< slice_header_disable_deblocking_filter_flag - uint8_t slice_loop_filter_across_slices_enabled_flag; - uint8_t collocated_list; - - unsigned int collocated_ref_idx; - - int slice_qp_delta; - int slice_cb_qp_offset; - int slice_cr_qp_offset; - - int beta_offset; ///< beta_offset_div2 * 2 - int tc_offset; ///< tc_offset_div2 * 2 - - unsigned int max_num_merge_cand; ///< 5 - 5_minus_max_num_merge_cand - - int num_entry_point_offsets; - - int8_t slice_qp; - - uint8_t luma_log2_weight_denom; - int16_t chroma_log2_weight_denom; - - int16_t luma_weight_l0[16]; - int16_t chroma_weight_l0[16][2]; - int16_t chroma_weight_l1[16][2]; - int16_t luma_weight_l1[16]; - - int16_t luma_offset_l0[16]; - int16_t chroma_offset_l0[16][2]; - - int16_t luma_offset_l1[16]; - int16_t chroma_offset_l1[16][2]; - - int slice_ctb_addr_rs; -} SliceHeader; - -typedef struct CodingTree { - int depth; ///< ctDepth -} CodingTree; - -typedef struct CodingUnit { - int x; - int y; - - enum PredMode pred_mode; ///< PredMode - enum PartMode part_mode; ///< PartMode - - uint8_t rqt_root_cbf; - - uint8_t pcm_flag; - - // Inferred parameters - uint8_t intra_split_flag; ///< IntraSplitFlag - uint8_t max_trafo_depth; ///< MaxTrafoDepth - uint8_t cu_transquant_bypass_flag; -} CodingUnit; - -typedef struct Mv { - int16_t x; ///< horizontal component of motion vector - int16_t y; ///< vertical component of motion vector -} Mv; - -typedef struct MvField { - Mv mv[2]; - int8_t ref_idx[2]; - int8_t pred_flag[2]; - uint8_t is_intra; -} MvField; - -typedef struct NeighbourAvailable { - int cand_bottom_left; - int cand_left; - int cand_up; - int cand_up_left; - int cand_up_right; - int cand_up_right_sap; -} NeighbourAvailable; - -typedef struct PredictionUnit { - int mpm_idx; - int rem_intra_luma_pred_mode; - uint8_t intra_pred_mode[4]; - Mv mvd; - uint8_t merge_flag; - uint8_t intra_pred_mode_c; -} PredictionUnit; - -typedef struct TransformTree { - uint8_t cbf_cb[MAX_TRANSFORM_DEPTH][MAX_CU_SIZE * MAX_CU_SIZE]; - uint8_t cbf_cr[MAX_TRANSFORM_DEPTH][MAX_CU_SIZE * MAX_CU_SIZE]; - uint8_t cbf_luma; - - // Inferred parameters - uint8_t inter_split_flag; -} TransformTree; - -typedef struct TransformUnit { - int cu_qp_delta; - - // Inferred parameters; - int cur_intra_pred_mode; - uint8_t is_cu_qp_delta_coded; -} TransformUnit; - -typedef struct SAOParams { - int offset_abs[3][4]; ///< sao_offset_abs - int offset_sign[3][4]; ///< sao_offset_sign - - int band_position[3]; ///< sao_band_position - - int eo_class[3]; ///< sao_eo_class - - int offset_val[3][5]; ///