aom/vp9/common/vp9_blockd.h

315 строки
8.7 KiB
C
Исходник Обычный вид История

2010-05-18 19:58:33 +04:00
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
2010-05-18 19:58:33 +04:00
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
2010-05-18 19:58:33 +04:00
*/
#ifndef VP9_COMMON_VP9_BLOCKD_H_
#define VP9_COMMON_VP9_BLOCKD_H_
2010-05-18 19:58:33 +04:00
#include "./vpx_config.h"
#include "vpx_ports/mem.h"
2010-05-18 19:58:33 +04:00
#include "vpx_scale/yv12config.h"
#include "vp9/common/vp9_common_data.h"
#include "vp9/common/vp9_filter.h"
#include "vp9/common/vp9_mv.h"
#include "vp9/common/vp9_scale.h"
2010-05-18 19:58:33 +04:00
#ifdef __cplusplus
extern "C" {
#endif
#define BLOCK_SIZE_GROUPS 4
#define SKIP_CONTEXTS 3
#define INTER_MODE_CONTEXTS 7
/* Segment Feature Masks */
#define MAX_MV_REF_CANDIDATES 2
2010-05-18 19:58:33 +04:00
#define INTRA_INTER_CONTEXTS 4
#define COMP_INTER_CONTEXTS 5
#define REF_CONTEXTS 5
typedef enum {
PLANE_TYPE_Y = 0,
PLANE_TYPE_UV = 1,
PLANE_TYPES
} PLANE_TYPE;
#define MAX_MB_PLANE 3
typedef char ENTROPY_CONTEXT;
2010-05-18 19:58:33 +04:00
static INLINE int combine_entropy_contexts(ENTROPY_CONTEXT a,
ENTROPY_CONTEXT b) {
return (a != 0) + (b != 0);
}
2010-05-18 19:58:33 +04:00
typedef enum {
KEY_FRAME = 0,
INTER_FRAME = 1,
FRAME_TYPES,
2010-05-18 19:58:33 +04:00
} FRAME_TYPE;
typedef enum {
DC_PRED, // Average of above and left pixels
V_PRED, // Vertical
H_PRED, // Horizontal
D45_PRED, // Directional 45 deg = round(arctan(1/1) * 180/pi)
D135_PRED, // Directional 135 deg = 180 - 45
D117_PRED, // Directional 117 deg = 180 - 63
D153_PRED, // Directional 153 deg = 180 - 27
D207_PRED, // Directional 207 deg = 180 + 27
D63_PRED, // Directional 63 deg = round(arctan(2/1) * 180/pi)
TM_PRED, // True-motion
NEARESTMV,
NEARMV,
ZEROMV,
NEWMV,
MB_MODE_COUNT
} PREDICTION_MODE;
2010-05-18 19:58:33 +04:00
static INLINE int is_inter_mode(PREDICTION_MODE mode) {
return mode >= NEARESTMV && mode <= NEWMV;
}
#define INTRA_MODES (TM_PRED + 1)
2010-05-18 19:58:33 +04:00
#define INTER_MODES (1 + NEWMV - NEARESTMV)
2010-05-18 19:58:33 +04:00
#define INTER_OFFSET(mode) ((mode) - NEARESTMV)
2010-05-18 19:58:33 +04:00
/* For keyframes, intra block modes are predicted by the (already decoded)
modes for the Y blocks to the left and above us; for interframes, there
is a single probability table. */
typedef struct {
PREDICTION_MODE as_mode;
int_mv as_mv[2]; // first, second inter predictor motion vectors
} b_mode_info;
2010-05-18 19:58:33 +04:00
// Note that the rate-distortion optimization loop, bit-stream writer, and
// decoder implementation modules critically rely on the enum entry values
// specified herein. They should be refactored concurrently.
typedef enum {
NONE = -1,
INTRA_FRAME = 0,
LAST_FRAME = 1,
GOLDEN_FRAME = 2,
ALTREF_FRAME = 3,
MAX_REF_FRAMES = 4
2010-05-18 19:58:33 +04:00
} MV_REFERENCE_FRAME;
// This structure now relates to 8x8 block regions.
typedef struct {
// Common for both INTER and INTRA blocks
BLOCK_SIZE sb_type;
PREDICTION_MODE mode;
TX_SIZE tx_size;
int8_t skip;
int8_t segment_id;
int8_t seg_id_predicted; // valid only when temporal_update is enabled
// Only for INTRA blocks
PREDICTION_MODE uv_mode;
// Only for INTER blocks
MV_REFERENCE_FRAME ref_frame[2];
int_mv mv[2];
int_mv ref_mvs[MAX_REF_FRAMES][MAX_MV_REF_CANDIDATES];
uint8_t mode_context[MAX_REF_FRAMES];
INTERP_FILTER interp_filter;
2010-05-18 19:58:33 +04:00
} MB_MODE_INFO;
typedef struct MODE_INFO {
struct MODE_INFO *src_mi;
MB_MODE_INFO mbmi;
b_mode_info bmi[4];
2010-05-18 19:58:33 +04:00
} MODE_INFO;
static INLINE PREDICTION_MODE get_y_mode(const MODE_INFO *mi, int block) {
return mi->mbmi.sb_type < BLOCK_8X8 ? mi->bmi[block].as_mode
: mi->mbmi.mode;
}
static INLINE int is_inter_block(const MB_MODE_INFO *mbmi) {
return mbmi->ref_frame[0] > INTRA_FRAME;
}
static INLINE int has_second_ref(const MB_MODE_INFO *mbmi) {
return mbmi->ref_frame[1] > INTRA_FRAME;
}
PREDICTION_MODE vp9_left_block_mode(const MODE_INFO *cur_mi,
const MODE_INFO *left_mi, int b);
PREDICTION_MODE vp9_above_block_mode(const MODE_INFO *cur_mi,
const MODE_INFO *above_mi, int b);
enum mv_precision {
MV_PRECISION_Q3,
MV_PRECISION_Q4
};
struct buf_2d {
uint8_t *buf;
int stride;
};
struct macroblockd_plane {
tran_low_t *dqcoeff;
PLANE_TYPE plane_type;
int subsampling_x;
int subsampling_y;
struct buf_2d dst;
struct buf_2d pre[2];
const int16_t *dequant;
ENTROPY_CONTEXT *above_context;
ENTROPY_CONTEXT *left_context;
};
#define BLOCK_OFFSET(x, i) ((x) + (i) * 16)
typedef struct RefBuffer {
// TODO(dkovalev): idx is not really required and should be removed, now it
// is used in vp9_onyxd_if.c
int idx;
YV12_BUFFER_CONFIG *buf;
struct scale_factors sf;
} RefBuffer;
typedef struct macroblockd {
struct macroblockd_plane plane[MAX_MB_PLANE];
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-08 02:45:05 +04:00
int mi_stride;
2010-05-18 19:58:33 +04:00
MODE_INFO *mi;
int up_available;
int left_available;
2010-05-18 19:58:33 +04:00
/* Distance of MB away from frame edges */
int mb_to_left_edge;
int mb_to_right_edge;
int mb_to_top_edge;
int mb_to_bottom_edge;
/* pointers to reference frames */
RefBuffer *block_refs[2];
/* pointer to current frame */
const YV12_BUFFER_CONFIG *cur_buf;
/* mc buffer */
DECLARE_ALIGNED(16, uint8_t, mc_buf[80 * 2 * 80 * 2]);
#if CONFIG_VP9_HIGHBITDEPTH
/* Bit depth: 8, 10, 12 */
int bd;
DECLARE_ALIGNED(16, uint16_t, mc_buf_high[80 * 2 * 80 * 2]);
#endif
int lossless;
int corrupted;
DECLARE_ALIGNED(16, tran_low_t, dqcoeff[MAX_MB_PLANE][64 * 64]);
ENTROPY_CONTEXT *above_context[MAX_MB_PLANE];
ENTROPY_CONTEXT left_context[MAX_MB_PLANE][16];
PARTITION_CONTEXT *above_seg_context;
PARTITION_CONTEXT left_seg_context[8];
2010-05-18 19:58:33 +04:00
} MACROBLOCKD;
static INLINE BLOCK_SIZE get_subsize(BLOCK_SIZE bsize,
PARTITION_TYPE partition) {
return subsize_lookup[partition][bsize];
}
extern const TX_TYPE intra_mode_to_tx_type_lookup[INTRA_MODES];
static INLINE TX_TYPE get_tx_type(PLANE_TYPE plane_type,
const MACROBLOCKD *xd) {
const MB_MODE_INFO *const mbmi = &xd->mi[0].src_mi->mbmi;
if (plane_type != PLANE_TYPE_Y || is_inter_block(mbmi))
return DCT_DCT;
return intra_mode_to_tx_type_lookup[mbmi->mode];
}
static INLINE TX_TYPE get_tx_type_4x4(PLANE_TYPE plane_type,
const MACROBLOCKD *xd, int ib) {
const MODE_INFO *const mi = xd->mi[0].src_mi;
if (plane_type != PLANE_TYPE_Y || xd->lossless || is_inter_block(&mi->mbmi))
return DCT_DCT;
return intra_mode_to_tx_type_lookup[get_y_mode(mi, ib)];
}
void vp9_setup_block_planes(MACROBLOCKD *xd, int ss_x, int ss_y);
static INLINE TX_SIZE get_uv_tx_size_impl(TX_SIZE y_tx_size, BLOCK_SIZE bsize,
int xss, int yss) {
if (bsize < BLOCK_8X8) {
return TX_4X4;
} else {
const BLOCK_SIZE plane_bsize = ss_size_lookup[bsize][xss][yss];
return MIN(y_tx_size, max_txsize_lookup[plane_bsize]);
}
}
2010-05-18 19:58:33 +04:00
static INLINE TX_SIZE get_uv_tx_size(const MB_MODE_INFO *mbmi,
const struct macroblockd_plane *pd) {
return get_uv_tx_size_impl(mbmi->tx_size, mbmi->sb_type, pd->subsampling_x,
pd->subsampling_y);
}
static INLINE BLOCK_SIZE get_plane_block_size(BLOCK_SIZE bsize,
const struct macroblockd_plane *pd) {
return ss_size_lookup[bsize][pd->subsampling_x][pd->subsampling_y];
}
typedef void (*foreach_transformed_block_visitor)(int plane, int block,
BLOCK_SIZE plane_bsize,
TX_SIZE tx_size,
void *arg);
void vp9_foreach_transformed_block_in_plane(
const MACROBLOCKD *const xd, BLOCK_SIZE bsize, int plane,
foreach_transformed_block_visitor visit, void *arg);
void vp9_foreach_transformed_block(
const MACROBLOCKD* const xd, BLOCK_SIZE bsize,
foreach_transformed_block_visitor visit, void *arg);
static INLINE void txfrm_block_to_raster_xy(BLOCK_SIZE plane_bsize,
TX_SIZE tx_size, int block,
int *x, int *y) {
const int bwl = b_width_log2_lookup[plane_bsize];
const int tx_cols_log2 = bwl - tx_size;
const int tx_cols = 1 << tx_cols_log2;
const int raster_mb = block >> (tx_size << 1);
*x = (raster_mb & (tx_cols - 1)) << tx_size;
*y = (raster_mb >> tx_cols_log2) << tx_size;
}
void vp9_set_contexts(const MACROBLOCKD *xd, struct macroblockd_plane *pd,
BLOCK_SIZE plane_bsize, TX_SIZE tx_size, int has_eob,
int aoff, int loff);
#ifdef __cplusplus
} // extern "C"
#endif
#endif // VP9_COMMON_VP9_BLOCKD_H_