4 * Copyright (C) 2012 - 2013 Guillaume Martres
6 * This file is part of Libav.
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #ifndef AVCODEC_HEVCDEC_H
24 #define AVCODEC_HEVCDEC_H
29 #include "libavutil/buffer.h"
30 #include "libavutil/md5.h"
36 #include "h2645_parse.h"
45 //TODO: check if this is really the maximum
46 #define MAX_TRANSFORM_DEPTH 5
48 #define MAX_TB_SIZE 32
49 #define MAX_PB_SIZE 64
51 #define DEFAULT_INTRA_TC_OFFSET 2
53 #define HEVC_CONTEXTS 183
55 #define MRG_MAX_NUM_CANDS 5
60 #define EPEL_EXTRA_BEFORE 1
61 #define EPEL_EXTRA_AFTER 2
64 #define EDGE_EMU_BUFFER_STRIDE 80
67 * Value of the luma sample at position (x, y) in the 2D array tab.
69 #define SAMPLE(tab, x, y) ((tab)[(y) * s->sps->width + (x)])
70 #define SAMPLE_CTB(tab, x, y) ((tab)[(y) * min_cb_width + (x)])
72 #define IS_IDR(s) (s->nal_unit_type == HEVC_NAL_IDR_W_RADL || s->nal_unit_type == HEVC_NAL_IDR_N_LP)
73 #define IS_BLA(s) (s->nal_unit_type == HEVC_NAL_BLA_W_RADL || s->nal_unit_type == HEVC_NAL_BLA_W_LP || \
74 s->nal_unit_type == HEVC_NAL_BLA_N_LP)
75 #define IS_IRAP(s) (s->nal_unit_type >= 16 && s->nal_unit_type <= 23)
77 #define FFUDIV(a,b) (((a) > 0 ? (a) : (a) - (b) + 1) / (b))
78 #define FFUMOD(a,b) ((a) - (b) * FFUDIV(a,b))
97 SPLIT_CODING_UNIT_FLAG,
98 CU_TRANSQUANT_BYPASS_FLAG,
104 PREV_INTRA_LUMA_PRED_FLAG,
106 REM_INTRA_LUMA_PRED_MODE,
107 INTRA_CHROMA_PRED_MODE,
113 ABS_MVD_GREATER0_FLAG,
114 ABS_MVD_GREATER1_FLAG,
118 NO_RESIDUAL_DATA_FLAG,
119 SPLIT_TRANSFORM_FLAG,
123 LAST_SIGNIFICANT_COEFF_X_PREFIX,
124 LAST_SIGNIFICANT_COEFF_Y_PREFIX,
125 LAST_SIGNIFICANT_COEFF_X_SUFFIX,
126 LAST_SIGNIFICANT_COEFF_Y_SUFFIX,
127 SIGNIFICANT_COEFF_GROUP_FLAG,
128 SIGNIFICANT_COEFF_FLAG,
129 COEFF_ABS_LEVEL_GREATER1_FLAG,
130 COEFF_ABS_LEVEL_GREATER2_FLAG,
131 COEFF_ABS_LEVEL_REMAINING,
215 typedef struct LongTermRPS {
221 typedef struct RefPicList {
222 struct HEVCFrame *ref[HEVC_MAX_REFS];
223 int list[HEVC_MAX_REFS];
224 int isLongTerm[HEVC_MAX_REFS];
228 typedef struct RefPicListTab {
229 RefPicList refPicList[2];
232 typedef struct SliceHeader {
235 ///< address (in raster order) of the first block in the current slice segment
236 unsigned int slice_segment_addr;
237 ///< address (in raster order) of the first block in the current slice
238 unsigned int slice_addr;
240 enum HEVCSliceType slice_type;
242 int pic_order_cnt_lsb;
244 uint8_t first_slice_in_pic_flag;
245 uint8_t dependent_slice_segment_flag;
246 uint8_t pic_output_flag;
247 uint8_t colour_plane_id;
249 ///< RPS coded in the slice header itself is stored here
250 int short_term_ref_pic_set_sps_flag;
251 int short_term_ref_pic_set_size;
252 ShortTermRPS slice_rps;
253 const ShortTermRPS *short_term_rps;
254 int long_term_ref_pic_set_size;
255 LongTermRPS long_term_rps;
256 unsigned int list_entry_lx[2][32];
258 uint8_t rpl_modification_flag[2];
259 uint8_t no_output_of_prior_pics_flag;
260 uint8_t slice_temporal_mvp_enabled_flag;
262 unsigned int nb_refs[2];
264 uint8_t slice_sample_adaptive_offset_flag[3];
265 uint8_t mvd_l1_zero_flag;
267 uint8_t cabac_init_flag;
268 uint8_t disable_deblocking_filter_flag; ///< slice_header_disable_deblocking_filter_flag
269 uint8_t slice_loop_filter_across_slices_enabled_flag;
270 uint8_t collocated_list;
272 unsigned int collocated_ref_idx;
275 int slice_cb_qp_offset;
276 int slice_cr_qp_offset;
278 int beta_offset; ///< beta_offset_div2 * 2
279 int tc_offset; ///< tc_offset_div2 * 2
281 unsigned int max_num_merge_cand; ///< 5 - 5_minus_max_num_merge_cand
283 int num_entry_point_offsets;
287 uint8_t luma_log2_weight_denom;
288 int16_t chroma_log2_weight_denom;
290 int16_t luma_weight_l0[16];
291 int16_t chroma_weight_l0[16][2];
292 int16_t chroma_weight_l1[16][2];
293 int16_t luma_weight_l1[16];
295 int16_t luma_offset_l0[16];
296 int16_t chroma_offset_l0[16][2];
298 int16_t luma_offset_l1[16];
299 int16_t chroma_offset_l1[16][2];
301 int slice_ctb_addr_rs;
304 typedef struct CodingTree {
305 int depth; ///< ctDepth
308 typedef struct CodingUnit {
312 enum PredMode pred_mode; ///< PredMode
313 enum PartMode part_mode; ///< PartMode
315 // Inferred parameters
316 uint8_t intra_split_flag; ///< IntraSplitFlag
317 uint8_t max_trafo_depth; ///< MaxTrafoDepth
318 uint8_t cu_transquant_bypass_flag;
322 int16_t x; ///< horizontal component of motion vector
323 int16_t y; ///< vertical component of motion vector
326 typedef struct MvField {
327 DECLARE_ALIGNED(4, Mv, mv)[2];
333 typedef struct NeighbourAvailable {
334 int cand_bottom_left;
339 int cand_up_right_sap;
340 } NeighbourAvailable;
342 typedef struct PredictionUnit {
344 int rem_intra_luma_pred_mode;
345 uint8_t intra_pred_mode[4];
348 uint8_t intra_pred_mode_c;
351 typedef struct TransformUnit {
354 // Inferred parameters;
355 int cur_intra_pred_mode;
356 uint8_t is_cu_qp_delta_coded;
359 typedef struct DBParams {
364 #define HEVC_FRAME_FLAG_OUTPUT (1 << 0)
365 #define HEVC_FRAME_FLAG_SHORT_REF (1 << 1)
366 #define HEVC_FRAME_FLAG_LONG_REF (1 << 2)
368 typedef struct HEVCFrame {
372 RefPicList *refPicList;
373 RefPicListTab **rpl_tab;
376 struct HEVCFrame *collocated_ref;
378 AVBufferRef *tab_mvf_buf;
379 AVBufferRef *rpl_tab_buf;
380 AVBufferRef *rpl_buf;
382 AVBufferRef *hwaccel_priv_buf;
383 void *hwaccel_picture_private;
386 * A sequence counter, so that old frames are output first
392 * A combination of HEVC_FRAME_FLAG_*
399 typedef struct HEVCPredContext {
400 void (*intra_pred[4])(struct HEVCContext *s, int x0, int y0, int c_idx);
402 void (*pred_planar[4])(uint8_t *src, const uint8_t *top,
403 const uint8_t *left, ptrdiff_t stride);
404 void (*pred_dc)(uint8_t *src, const uint8_t *top, const uint8_t *left,
405 ptrdiff_t stride, int log2_size, int c_idx);
406 void (*pred_angular[4])(uint8_t *src, const uint8_t *top,
407 const uint8_t *left, ptrdiff_t stride,
408 int c_idx, int mode);
411 typedef struct HEVCLocalContext {
412 DECLARE_ALIGNED(16, int16_t, mc_buffer[(MAX_PB_SIZE + 24) * MAX_PB_SIZE]);
413 uint8_t cabac_state[HEVC_CONTEXTS];
415 uint8_t first_qp_group;
425 uint8_t ctb_left_flag;
427 uint8_t ctb_up_right_flag;
428 uint8_t ctb_up_left_flag;
429 int start_of_tiles_x;
432 /* +7 is for subpixel interpolation, *2 for high bit depths */
433 DECLARE_ALIGNED(32, uint8_t, edge_emu_buffer)[(MAX_PB_SIZE + 7) * EDGE_EMU_BUFFER_STRIDE * 2];
437 NeighbourAvailable na;
439 #define BOUNDARY_LEFT_SLICE (1 << 0)
440 #define BOUNDARY_LEFT_TILE (1 << 1)
441 #define BOUNDARY_UPPER_SLICE (1 << 2)
442 #define BOUNDARY_UPPER_TILE (1 << 3)
443 /* properties of the boundary of the current CTB for the purposes
444 * of the deblocking filter */
448 typedef struct HEVCContext {
449 const AVClass *c; // needed by private avoptions
450 AVCodecContext *avctx;
452 HEVCLocalContext HEVClc;
454 uint8_t cabac_state[HEVC_CONTEXTS];
456 /** 1 if the independent slice segment header was successfully parsed */
457 uint8_t slice_initialized;
462 AVFrame *output_frame;
466 struct AVMD5 *md5_ctx;
468 AVBufferPool *tab_mvf_pool;
469 AVBufferPool *rpl_tab_pool;
471 ///< candidate references for the current frame
477 enum HEVCNALUnitType nal_unit_type;
478 int temporal_id; ///< temporal_id_plus1 - 1
483 int slice_idx; ///< number of the slice being currently decoded
484 int eos; ///< current packet contains an EOS/EOB NAL
492 HEVCDSPContext hevcdsp;
493 VideoDSPContext vdsp;
494 BswapDSPContext bdsp;
496 uint8_t *horizontal_bs;
497 uint8_t *vertical_bs;
499 int32_t *tab_slice_address;
503 uint8_t *tab_ct_depth;
507 uint8_t *cbf_luma; // cbf_luma of colocated TU
510 // CTB-level flags affecting loop filter operation
511 uint8_t *filter_slice_edges;
513 /** used on BE to byteswap the lines for checksumming */
514 uint8_t *checksum_buf;
515 int checksum_buf_size;
518 * Sequence counters for decoded and output frames, so that old
519 * frames are output first after a POC reset
525 // type of the first VCL NAL of the current frame
526 enum HEVCNALUnitType first_nal_type;
528 uint8_t context_initialized;
529 uint8_t is_nalff; ///< this flag is != 0 if bitstream is encapsulated
530 ///< as a format defined in 14496-15
531 int apply_defdispwin;
533 int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
538 * Mark all frames in DPB as unused for reference.
540 void ff_hevc_clear_refs(HEVCContext *s);
543 * Drop all frames currently in DPB.
545 void ff_hevc_flush_dpb(HEVCContext *s);
548 * Compute POC of the current frame and return it.
550 int ff_hevc_compute_poc(HEVCContext *s, int poc_lsb);
552 RefPicList *ff_hevc_get_ref_list(HEVCContext *s, HEVCFrame *frame,
556 * Construct the reference picture sets for the current frame.
558 int ff_hevc_frame_rps(HEVCContext *s);
561 * Construct the reference picture list(s) for the current slice.
563 int ff_hevc_slice_rpl(HEVCContext *s);
565 void ff_hevc_save_states(HEVCContext *s, int ctb_addr_ts);
566 void ff_hevc_cabac_init(HEVCContext *s, int ctb_addr_ts);
567 int ff_hevc_sao_merge_flag_decode(HEVCContext *s);
568 int ff_hevc_sao_type_idx_decode(HEVCContext *s);
569 int ff_hevc_sao_band_position_decode(HEVCContext *s);
570 int ff_hevc_sao_offset_abs_decode(HEVCContext *s);
571 int ff_hevc_sao_offset_sign_decode(HEVCContext *s);
572 int ff_hevc_sao_eo_class_decode(HEVCContext *s);
573 int ff_hevc_end_of_slice_flag_decode(HEVCContext *s);
574 int ff_hevc_cu_transquant_bypass_flag_decode(HEVCContext *s);
575 int ff_hevc_skip_flag_decode(HEVCContext *s, int x0, int y0,
577 int ff_hevc_pred_mode_decode(HEVCContext *s);
578 int ff_hevc_split_coding_unit_flag_decode(HEVCContext *s, int ct_depth,
580 int ff_hevc_part_mode_decode(HEVCContext *s, int log2_cb_size);
581 int ff_hevc_pcm_flag_decode(HEVCContext *s);
582 int ff_hevc_prev_intra_luma_pred_flag_decode(HEVCContext *s);
583 int ff_hevc_mpm_idx_decode(HEVCContext *s);
584 int ff_hevc_rem_intra_luma_pred_mode_decode(HEVCContext *s);
585 int ff_hevc_intra_chroma_pred_mode_decode(HEVCContext *s);
586 int ff_hevc_merge_idx_decode(HEVCContext *s);
587 int ff_hevc_merge_flag_decode(HEVCContext *s);
588 int ff_hevc_inter_pred_idc_decode(HEVCContext *s, int nPbW, int nPbH);
589 int ff_hevc_ref_idx_lx_decode(HEVCContext *s, int num_ref_idx_lx);
590 int ff_hevc_mvp_lx_flag_decode(HEVCContext *s);
591 int ff_hevc_no_residual_syntax_flag_decode(HEVCContext *s);
592 int ff_hevc_abs_mvd_greater0_flag_decode(HEVCContext *s);
593 int ff_hevc_abs_mvd_greater1_flag_decode(HEVCContext *s);
594 int ff_hevc_mvd_decode(HEVCContext *s);
595 int ff_hevc_mvd_sign_flag_decode(HEVCContext *s);
596 int ff_hevc_split_transform_flag_decode(HEVCContext *s, int log2_trafo_size);
597 int ff_hevc_cbf_cb_cr_decode(HEVCContext *s, int trafo_depth);
598 int ff_hevc_cbf_luma_decode(HEVCContext *s, int trafo_depth);
599 int ff_hevc_transform_skip_flag_decode(HEVCContext *s, int c_idx);
600 int ff_hevc_last_significant_coeff_x_prefix_decode(HEVCContext *s, int c_idx,
602 int ff_hevc_last_significant_coeff_y_prefix_decode(HEVCContext *s, int c_idx,
604 int ff_hevc_last_significant_coeff_suffix_decode(HEVCContext *s,
605 int last_significant_coeff_prefix);
606 int ff_hevc_significant_coeff_group_flag_decode(HEVCContext *s, int c_idx,
608 int ff_hevc_significant_coeff_flag_decode(HEVCContext *s, int c_idx, int x_c,
609 int y_c, int log2_trafo_size,
610 int scan_idx, int prev_sig);
611 int ff_hevc_coeff_abs_level_greater1_flag_decode(HEVCContext *s, int c_idx,
613 int ff_hevc_coeff_abs_level_greater2_flag_decode(HEVCContext *s, int c_idx,
615 int ff_hevc_coeff_abs_level_remaining(HEVCContext *s, int base_level,
617 int ff_hevc_coeff_sign_flag(HEVCContext *s, uint8_t nb);
620 * Get the number of candidate references for the current frame.
622 int ff_hevc_frame_nb_refs(HEVCContext *s);
624 int ff_hevc_set_new_ref(HEVCContext *s, AVFrame **frame, int poc);
627 * Find next frame in output order and put a reference to it in frame.
628 * @return 1 if a frame was output, 0 otherwise
630 int ff_hevc_output_frame(HEVCContext *s, AVFrame *frame, int flush);
632 void ff_hevc_unref_frame(HEVCContext *s, HEVCFrame *frame, int flags);
634 void ff_hevc_set_neighbour_available(HEVCContext *s, int x0, int y0,
636 void ff_hevc_luma_mv_merge_mode(HEVCContext *s, int x0, int y0,
637 int nPbW, int nPbH, int log2_cb_size,
638 int part_idx, int merge_idx, MvField *mv);
639 void ff_hevc_luma_mv_mvp_mode(HEVCContext *s, int x0, int y0,
640 int nPbW, int nPbH, int log2_cb_size,
641 int part_idx, int merge_idx,
642 MvField *mv, int mvp_lx_flag, int LX);
643 void ff_hevc_set_qPy(HEVCContext *s, int xC, int yC, int xBase, int yBase,
645 void ff_hevc_deblocking_boundary_strengths(HEVCContext *s, int x0, int y0,
646 int log2_trafo_size);
647 int ff_hevc_cu_qp_delta_sign_flag(HEVCContext *s);
648 int ff_hevc_cu_qp_delta_abs(HEVCContext *s);
649 void ff_hevc_hls_filter(HEVCContext *s, int x, int y);
650 void ff_hevc_hls_filters(HEVCContext *s, int x_ctb, int y_ctb, int ctb_size);
652 void ff_hevc_pred_init(HEVCPredContext *hpc, int bit_depth);
654 extern const uint8_t ff_hevc_qpel_extra_before[4];
655 extern const uint8_t ff_hevc_qpel_extra_after[4];
656 extern const uint8_t ff_hevc_qpel_extra[4];
658 #endif /* AVCODEC_HEVCDEC_H */