mirror of
https://github.com/ultravideo/uvg266.git
synced 2024-11-27 19:24:06 +00:00
Refactor: filter.c/.h full cleanup.
This commit is contained in:
Ari Koivula
parent
61c1d7413f
commit
4416d6ec36
316
src/filter.c
316
src/filter.c
|
|
@ -19,21 +19,41 @@
|
||||||
#include "bitstream.h"
|
#include "bitstream.h"
|
||||||
#include "picture.h"
|
#include "picture.h"
|
||||||
#include "cabac.h"
|
#include "cabac.h"
|
||||||
|
#include "transform.h"
|
||||||
|
|
||||||
|
//////////////////////////////////////////////////////////////////////////
|
||||||
extern const uint8_t g_chroma_scale[58];
|
// INITIALIZATIONS
|
||||||
const uint8_t tctable_8x8[54] =
|
const uint8_t g_tc_table_8x8[54] =
|
||||||
{
|
{
|
||||||
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,5,5,6,6,7,8,9,10,11,13,14,16,18,20,22,24
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
||||||
|
0, 0, 0, 0, 0, 0, 0, 0, 1, 1,
|
||||||
|
1, 1, 1, 1, 1, 1, 1, 2, 2, 2,
|
||||||
|
2, 3, 3, 3, 3, 4, 4, 4, 5, 5,
|
||||||
|
6, 6, 7, 8, 9, 10, 11, 13, 14, 16,
|
||||||
|
18, 20, 22, 24
|
||||||
};
|
};
|
||||||
|
|
||||||
const uint8_t betatable_8x8[52] =
|
const uint8_t g_beta_table_8x8[52] =
|
||||||
{
|
{
|
||||||
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,6,7,8,9,10,11,12,13,14,15,16,17,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
||||||
|
0, 0, 0, 0, 0, 0, 6, 7, 8, 9,
|
||||||
|
10, 11, 12, 13, 14, 15, 16, 17, 18, 20,
|
||||||
|
22, 24, 26, 28, 30, 32, 34, 36, 38, 40,
|
||||||
|
42, 44, 46, 48, 50, 52, 54, 56, 58, 60,
|
||||||
|
62, 64
|
||||||
};
|
};
|
||||||
|
|
||||||
|
//////////////////////////////////////////////////////////////////////////
|
||||||
|
// FUNCTIONS
|
||||||
|
|
||||||
INLINE void filter_deblock_luma( uint8_t* src, int32_t offset, int32_t tc , int8_t sw, int8_t part_P_nofilter, int8_t part_Q_nofilter, int32_t thr_cut, int8_t filter_second_P, int8_t filter_second_Q)
|
/**
|
||||||
|
* \brief
|
||||||
|
*/
|
||||||
|
INLINE void filter_deblock_luma(uint8_t *src, int32_t offset,
|
||||||
|
int32_t tc, int8_t sw,
|
||||||
|
int8_t part_P_nofilter, int8_t part_Q_nofilter,
|
||||||
|
int32_t thr_cut,
|
||||||
|
int8_t filter_second_P, int8_t filter_second_Q)
|
||||||
{
|
{
|
||||||
int32_t delta;
|
int32_t delta;
|
||||||
|
|
||||||
|
|
@ -46,114 +66,110 @@ INLINE void filter_deblock_luma( uint8_t* src, int32_t offset, int32_t tc , int8
|
||||||
int16_t m6 = src[offset * 2];
|
int16_t m6 = src[offset * 2];
|
||||||
int16_t m7 = src[offset * 3];
|
int16_t m7 = src[offset * 3];
|
||||||
|
|
||||||
if (sw)
|
if (sw) {
|
||||||
{
|
src[-offset * 3] = CLIP(m1 - 2*tc, m1 + 2*tc, (2*m0 + 3*m1 + m2 + m3 + m4 + 4) >> 3);
|
||||||
src[-offset*3] = CLIP(m1-2*tc, m1+2*tc, ((2*m0 + 3*m1 + m2 + m3 + m4 + 4 )>>3));
|
src[-offset * 2] = CLIP(m2 - 2*tc, m2 + 2*tc, ( m1 + m2 + m3 + m4 + 2) >> 2);
|
||||||
src[-offset*2] = CLIP(m2-2*tc, m2+2*tc, ((m1 + m2 + m3 + m4 + 2)>>2));
|
src[-offset] = CLIP(m3 - 2*tc, m3 + 2*tc, ( m1 + 2*m2 + 2*m3 + 2*m4 + m5 + 4) >> 3);
|
||||||
src[-offset] = CLIP(m3-2*tc, m3+2*tc, ((m1 + 2*m2 + 2*m3 + 2*m4 + m5 + 4) >> 3));
|
src[0] = CLIP(m4 - 2*tc, m4 + 2*tc, ( m2 + 2*m3 + 2*m4 + 2*m5 + m6 + 4) >> 3);
|
||||||
src[0] = CLIP(m4-2*tc, m4+2*tc, ((m2 + 2*m3 + 2*m4 + 2*m5 + m6 + 4) >> 3));
|
src[offset] = CLIP(m5 - 2*tc, m5 + 2*tc, ( m3 + m4 + m5 + m6 + 2) >> 2);
|
||||||
src[ offset] = CLIP(m5-2*tc, m5+2*tc, ((m3 + m4 + m5 + m6 + 2)>>2));
|
src[offset * 2] = CLIP(m6 - 2*tc, m6 + 2*tc, ( m3 + m4 + m5 + 3*m6 + 2*m7 + 4) >> 3);
|
||||||
src[ offset*2] = CLIP(m6-2*tc, m6+2*tc, ((m3 + m4 + m5 + 3*m6 + 2*m7 + 4 )>>3));
|
} else {
|
||||||
}
|
|
||||||
else
|
|
||||||
{
|
|
||||||
// Weak filter
|
// Weak filter
|
||||||
delta = (9*(m4 - m3) - 3*(m5 - m2) + 8) >> 4;
|
delta = (9*(m4 - m3) - 3*(m5 - m2) + 8) >> 4;
|
||||||
|
|
||||||
if ( abs(delta) < thr_cut )
|
if (abs(delta) < thr_cut) {
|
||||||
{
|
|
||||||
int32_t tc2 = tc >> 1;
|
int32_t tc2 = tc >> 1;
|
||||||
delta = CLIP(-tc, tc, delta);
|
delta = CLIP(-tc, tc, delta);
|
||||||
src[-offset] = CLIP(0, (1 << g_bitdepth) - 1, (m3 + delta));
|
src[-offset] = CLIP(0, (1 << g_bitdepth) - 1, (m3 + delta));
|
||||||
src[0] = CLIP(0, (1 << g_bitdepth) - 1, (m4 - delta));
|
src[0] = CLIP(0, (1 << g_bitdepth) - 1, (m4 - delta));
|
||||||
|
|
||||||
|
if(filter_second_P) {
|
||||||
if(filter_second_P)
|
int32_t delta1 = CLIP(-tc2, tc2, (((m1 + m3 + 1) >> 1) - m2 + delta) >> 1);
|
||||||
{
|
src[-offset * 2] = CLIP(0, (1 << g_bitdepth) - 1, m2 + delta1);
|
||||||
int32_t delta1 = CLIP(-tc2, tc2, (( ((m1+m3+1)>>1)- m2+delta)>>1));
|
|
||||||
src[-offset*2] = CLIP(0,(1 << g_bitdepth)-1,(m2+delta1));
|
|
||||||
}
|
}
|
||||||
if(filter_second_Q)
|
if(filter_second_Q) {
|
||||||
{
|
int32_t delta2 = CLIP(-tc2, tc2, (((m6 + m4 + 1) >> 1) - m5 - delta) >> 1);
|
||||||
int32_t delta2 = CLIP(-tc2, tc2, (( ((m6+m4+1)>>1)- m5-delta)>>1));
|
src[offset] = CLIP(0, (1 << g_bitdepth) - 1, m5 + delta2);
|
||||||
src[ offset] = CLIP(0,(1 << g_bitdepth)-1,(m5+delta2));
|
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
if(part_P_nofilter)
|
if(part_P_nofilter) {
|
||||||
{
|
|
||||||
src[-offset] = (uint8_t)m3;
|
src[-offset] = (uint8_t)m3;
|
||||||
src[-offset*2] = (uint8_t)m2;
|
src[-offset*2] = (uint8_t)m2;
|
||||||
src[-offset*3] = (uint8_t)m1;
|
src[-offset*3] = (uint8_t)m1;
|
||||||
}
|
}
|
||||||
if(part_Q_nofilter)
|
if(part_Q_nofilter) {
|
||||||
{
|
|
||||||
src[0] = (uint8_t)m4;
|
src[0] = (uint8_t)m4;
|
||||||
src[offset] = (uint8_t)m5;
|
src[offset] = (uint8_t)m5;
|
||||||
src[offset*2] = (uint8_t)m6;
|
src[offset*2] = (uint8_t)m6;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
INLINE void filter_deblock_chroma( uint8_t* src, int32_t offset, int32_t tc ,int8_t part_P_nofilter, int8_t part_Q_nofilter)
|
/**
|
||||||
|
* \brief
|
||||||
|
*/
|
||||||
|
INLINE void filter_deblock_chroma(uint8_t *src, int32_t offset, int32_t tc,
|
||||||
|
int8_t part_P_nofilter, int8_t part_Q_nofilter)
|
||||||
{
|
{
|
||||||
int32_t delta;
|
int32_t delta;
|
||||||
|
|
||||||
int16_t m2 = src[-offset * 2];
|
int16_t m2 = src[-offset * 2];
|
||||||
int16_t m3 = src[-offset];
|
int16_t m3 = src[-offset];
|
||||||
int16_t m4 = src[0];
|
int16_t m4 = src[0];
|
||||||
int16_t m5 = src[offset];
|
int16_t m5 = src[offset];
|
||||||
|
|
||||||
delta = CLIP(-tc,tc, (((( m4 - m3 ) << 2 ) + m2 - m5 + 4 ) >> 3) );
|
delta = CLIP(-tc,tc, (((m4 - m3) << 2) + m2 - m5 + 4 ) >> 3);
|
||||||
if(!part_P_nofilter)
|
if(!part_P_nofilter) {
|
||||||
{
|
|
||||||
src[-offset] = CLIP(0, (1 << g_bitdepth) - 1, m3 + delta);
|
src[-offset] = CLIP(0, (1 << g_bitdepth) - 1, m3 + delta);
|
||||||
}
|
}
|
||||||
if(!part_Q_nofilter)
|
if(!part_Q_nofilter) {
|
||||||
{
|
|
||||||
src[0] = CLIP(0, (1 << g_bitdepth) - 1, m4 - delta);
|
src[0] = CLIP(0, (1 << g_bitdepth) - 1, m4 - delta);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t ypos, int8_t depth, int8_t dir)
|
/**
|
||||||
|
* \brief
|
||||||
|
*/
|
||||||
|
void filter_deblock_edge_luma(encoder_control *encoder,
|
||||||
|
int32_t xpos, int32_t ypos,
|
||||||
|
int8_t depth, int8_t dir)
|
||||||
{
|
{
|
||||||
int32_t stride = encoder->in.cur_pic->width;
|
int32_t stride = encoder->in.cur_pic->width;
|
||||||
int32_t offset = stride;
|
int32_t offset = stride;
|
||||||
int32_t betaOffsetDiv2 = encoder->beta_offset_div2;
|
int32_t beta_offset_div2 = encoder->beta_offset_div2;
|
||||||
int32_t tcOffsetDiv2 = encoder->tc_offset_div2;
|
int32_t tc_offset_div2 = encoder->tc_offset_div2;
|
||||||
int8_t uiBs = 2; /* Filter strength */
|
int8_t strength = 2; // Filter strength
|
||||||
/* TODO: support 10+bits */
|
// TODO: support 10+bits
|
||||||
uint8_t* origsrc = &encoder->in.cur_pic->y_recdata[xpos+ypos*stride];
|
uint8_t *orig_src = &encoder->in.cur_pic->y_recdata[xpos + ypos*stride];
|
||||||
uint8_t* src = origsrc;
|
uint8_t *src = orig_src;
|
||||||
int32_t step = 1;
|
int32_t step = 1;
|
||||||
//CU_info* cu = &encoder->in.cur_pic->CU[depth][(xpos>>scu_width_log2) + (ypos>>scu_width_log2)*(encoder->in.width>>scu_width_log2)];
|
|
||||||
|
|
||||||
if(dir == EDGE_VER)
|
if(dir == EDGE_VER) {
|
||||||
{
|
|
||||||
offset = 1;
|
offset = 1;
|
||||||
step = stride;
|
step = stride;
|
||||||
}
|
}
|
||||||
|
|
||||||
{
|
{
|
||||||
int32_t QP = encoder->QP;
|
int32_t qp = encoder->QP;
|
||||||
int32_t bitdepth_scale = 1 << (g_bitdepth - 8);
|
int32_t bitdepth_scale = 1 << (g_bitdepth - 8);
|
||||||
int32_t TC_index = CLIP(0, 51+2, (int32_t)(QP + 2*(uiBs-1) + (tcOffsetDiv2 << 1)));
|
int32_t tc_index = CLIP(0, 51 + 2, (int32_t)(qp + 2*(strength - 1) + (tc_offset_div2 << 1)));
|
||||||
int32_t B_index = CLIP(0, 51, QP + (betaOffsetDiv2 << 1));
|
int32_t b_index = CLIP(0, 51, qp + (beta_offset_div2 << 1));
|
||||||
int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale;
|
int32_t tc = g_tc_table_8x8[tc_index] * bitdepth_scale;
|
||||||
int32_t Beta = betatable_8x8[B_index]*bitdepth_scale;
|
int32_t beta = g_beta_table_8x8[b_index] * bitdepth_scale;
|
||||||
int32_t side_threshold = (Beta+(Beta>>1))>>3;
|
int32_t side_threshold = (beta + (beta >>1 )) >> 3;
|
||||||
int32_t iThrCut = Tc*10;
|
int32_t thr_cut = tc * 10;
|
||||||
uint32_t blocks_in_part = (LCU_WIDTH >> depth) / 4;
|
uint32_t blocks_in_part = (LCU_WIDTH >> depth) / 4;
|
||||||
uint32_t block_idx;
|
uint32_t block_idx;
|
||||||
|
|
||||||
/* TODO: add CU based QP calculation */
|
// TODO: add CU based QP calculation
|
||||||
|
|
||||||
/* For each 4-pixel part in the edge */
|
// For each 4-pixel part in the edge
|
||||||
for (block_idx = 0; block_idx < blocks_in_part; block_idx++)
|
for (block_idx = 0; block_idx < blocks_in_part; ++block_idx)
|
||||||
{
|
{
|
||||||
int32_t dp0, dq0, dp3, dq3, d0, d3, dp, dq, d;
|
int32_t dp0, dq0, dp3, dq3, d0, d3, dp, dq, d;
|
||||||
|
|
||||||
/* Check conditions for filtering */
|
// Check conditions for filtering
|
||||||
|
// TODO: Get rid of these inline defines.
|
||||||
#define calc_DP(s,o) abs( (int16_t)s[-o*3] - (int16_t)2*s[-o*2] + (int16_t)s[-o] )
|
#define calc_DP(s,o) abs( (int16_t)s[-o*3] - (int16_t)2*s[-o*2] + (int16_t)s[-o] )
|
||||||
#define calc_DQ(s,o) abs( (int16_t)s[0] - (int16_t)2*s[o] + (int16_t)s[o*2] )
|
#define calc_DQ(s,o) abs( (int16_t)s[0] - (int16_t)2*s[o] + (int16_t)s[o*2] )
|
||||||
|
|
||||||
|
|
@ -167,43 +183,48 @@ void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t yp
|
||||||
dq = dq0 + dq3;
|
dq = dq0 + dq3;
|
||||||
d = d0 + d3;
|
d = d0 + d3;
|
||||||
|
|
||||||
#if ENABLE_PCM == 1
|
#if ENABLE_PCM
|
||||||
// TODO: add PCM deblocking
|
// TODO: add PCM deblocking
|
||||||
#endif
|
#endif
|
||||||
if (d < Beta)
|
|
||||||
{
|
if (d < beta) {
|
||||||
int8_t filter_P = (dp < side_threshold) ? 1 : 0;
|
int8_t filter_P = (dp < side_threshold) ? 1 : 0;
|
||||||
int8_t filter_Q = (dq < side_threshold) ? 1 : 0;
|
int8_t filter_Q = (dq < side_threshold) ? 1 : 0;
|
||||||
|
|
||||||
/* Strong filtering flag checking */
|
// Strong filtering flag checking
|
||||||
#define useStrongFiltering(o,d,s) ( ((abs(s[-o*4]-s[-o]) + abs(s[o*3]-s[0])) < (Beta>>3)) && (d<(Beta>>2)) && ( abs(s[-o]-s[0]) < ((Tc*5+1)>>1)) )
|
#define useStrongFiltering(o,d,s) ( ((abs(s[-o*4]-s[-o]) + abs(s[o*3]-s[0])) < (beta>>3)) && (d<(beta>>2)) && ( abs(s[-o]-s[0]) < ((tc*5+1)>>1)) )
|
||||||
int8_t sw = useStrongFiltering(offset, 2*d0, (src+step*(block_idx*4+0))) &&
|
int8_t sw = useStrongFiltering(offset, 2*d0, (src+step*(block_idx*4+0))) &&
|
||||||
useStrongFiltering(offset, 2*d3, (src+step*(block_idx*4+3)));
|
useStrongFiltering(offset, 2*d3, (src+step*(block_idx*4+3)));
|
||||||
|
|
||||||
/* Filter four rows/columns */
|
// Filter four rows/columns
|
||||||
filter_deblock_luma( src+step*(block_idx*4+0), offset, Tc, sw, 0, 0, iThrCut, filter_P, filter_Q);
|
filter_deblock_luma(src + step * (4*block_idx + 0), offset, tc, sw, 0, 0, thr_cut, filter_P, filter_Q);
|
||||||
filter_deblock_luma( src+step*(block_idx*4+1), offset, Tc, sw, 0, 0, iThrCut, filter_P, filter_Q);
|
filter_deblock_luma(src + step * (4*block_idx + 1), offset, tc, sw, 0, 0, thr_cut, filter_P, filter_Q);
|
||||||
filter_deblock_luma( src+step*(block_idx*4+2), offset, Tc, sw, 0, 0, iThrCut, filter_P, filter_Q);
|
filter_deblock_luma(src + step * (4*block_idx + 2), offset, tc, sw, 0, 0, thr_cut, filter_P, filter_Q);
|
||||||
filter_deblock_luma( src+step*(block_idx*4+3), offset, Tc, sw, 0, 0, iThrCut, filter_P, filter_Q);
|
filter_deblock_luma(src + step * (4*block_idx + 3), offset, tc, sw, 0, 0, thr_cut, filter_P, filter_Q);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t ypos, int8_t depth, int8_t dir)
|
/**
|
||||||
|
* \brief
|
||||||
|
*/
|
||||||
|
void filter_deblock_edge_chroma(encoder_control *encoder,
|
||||||
|
int32_t x, int32_t y,
|
||||||
|
int8_t depth, int8_t dir)
|
||||||
{
|
{
|
||||||
int32_t stride = encoder->in.cur_pic->width >> 1;
|
int32_t stride = encoder->in.cur_pic->width >> 1;
|
||||||
int32_t tcOffsetDiv2 = encoder->tc_offset_div2;
|
int32_t tc_offset_div2 = encoder->tc_offset_div2;
|
||||||
int8_t uiNumParts = 1;
|
int8_t num_parts = 1;
|
||||||
/* TODO: support 10+bits */
|
// TODO: support 10+bits
|
||||||
uint8_t* srcU = &encoder->in.cur_pic->u_recdata[xpos+ypos*stride];
|
uint8_t* src_u = &encoder->in.cur_pic->u_recdata[x + y*stride];
|
||||||
uint8_t* srcV = &encoder->in.cur_pic->v_recdata[xpos+ypos*stride];
|
uint8_t* src_v = &encoder->in.cur_pic->v_recdata[x + y*stride];
|
||||||
/* Init offset and step to EDGE_HOR */
|
// Init offset and step to EDGE_HOR
|
||||||
int32_t offset = stride;
|
int32_t offset = stride;
|
||||||
int32_t step = 1;
|
int32_t step = 1;
|
||||||
|
|
||||||
/* We cannot filter edges not on 8x8 grid */
|
// We cannot filter edges not on 8x8 grid
|
||||||
if( depth == MAX_DEPTH && (( (ypos & 0x7) && dir == EDGE_HOR ) || ( (xpos & 0x7) && dir == EDGE_VER ) ) )
|
if(depth == MAX_DEPTH && (( (y & 0x7) && dir == EDGE_HOR ) || ( (x & 0x7) && dir == EDGE_VER ) ) )
|
||||||
{
|
{
|
||||||
return;
|
return;
|
||||||
}
|
}
|
||||||
|
|
@ -218,108 +239,103 @@ void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t y
|
||||||
{
|
{
|
||||||
int32_t QP = g_chroma_scale[encoder->QP];
|
int32_t QP = g_chroma_scale[encoder->QP];
|
||||||
int32_t bitdepth_scale = 1 << (g_bitdepth-8);
|
int32_t bitdepth_scale = 1 << (g_bitdepth-8);
|
||||||
int32_t TC_index = CLIP(0, 51+2, (int32_t)(QP + 2 + (tcOffsetDiv2 << 1)));
|
int32_t TC_index = CLIP(0, 51+2, (int32_t)(QP + 2 + (tc_offset_div2 << 1)));
|
||||||
int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale;
|
int32_t Tc = g_tc_table_8x8[TC_index]*bitdepth_scale;
|
||||||
uint32_t blocks_in_part= (LCU_WIDTH>>(depth+1)) / 4;
|
uint32_t blocks_in_part= (LCU_WIDTH>>(depth+1)) / 4;
|
||||||
uint32_t blk_idx;
|
uint32_t blk_idx;
|
||||||
|
|
||||||
for (blk_idx = 0; blk_idx < blocks_in_part; blk_idx++)
|
for (blk_idx = 0; blk_idx < blocks_in_part; ++blk_idx)
|
||||||
{
|
{
|
||||||
/* Chroma U */
|
// Chroma U
|
||||||
filter_deblock_chroma( srcU+step*(blk_idx*4+0), offset, Tc,0, 0);
|
filter_deblock_chroma(src_u + step * (4*blk_idx + 0), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcU+step*(blk_idx*4+1), offset, Tc,0, 0);
|
filter_deblock_chroma(src_u + step * (4*blk_idx + 1), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcU+step*(blk_idx*4+2), offset, Tc,0, 0);
|
filter_deblock_chroma(src_u + step * (4*blk_idx + 2), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcU+step*(blk_idx*4+3), offset, Tc,0, 0);
|
filter_deblock_chroma(src_u + step * (4*blk_idx + 3), offset, Tc, 0, 0);
|
||||||
/* Chroma V */
|
// Chroma V
|
||||||
filter_deblock_chroma( srcV+step*(blk_idx*4+0), offset, Tc,0, 0);
|
filter_deblock_chroma(src_v + step * (4*blk_idx + 0), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcV+step*(blk_idx*4+1), offset, Tc,0, 0);
|
filter_deblock_chroma(src_v + step * (4*blk_idx + 1), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcV+step*(blk_idx*4+2), offset, Tc,0, 0);
|
filter_deblock_chroma(src_v + step * (4*blk_idx + 2), offset, Tc, 0, 0);
|
||||||
filter_deblock_chroma( srcV+step*(blk_idx*4+3), offset, Tc,0, 0);
|
filter_deblock_chroma(src_v + step * (4*blk_idx + 3), offset, Tc, 0, 0);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
/*! \brief function to split LCU into smaller CU blocks
|
/**
|
||||||
\param encoder the encoder info structure
|
* \brief function to split LCU into smaller CU blocks
|
||||||
\param xCtb block x-position (as SCU)
|
* \param encoder the encoder info structure
|
||||||
\param yCtb block y-position (as SCU)
|
* \param xCtb block x-position (as SCU)
|
||||||
\param depth block depth
|
* \param yCtb block y-position (as SCU)
|
||||||
\param edge which edge we are filtering
|
* \param depth block depth
|
||||||
|
* \param edge which edge we are filtering
|
||||||
This function takes (SCU) block position as input and splits the block
|
*
|
||||||
until the coded block size has been achived. Calls luma and chroma filtering
|
* This function takes (SCU) block position as input and splits the block
|
||||||
functions for each coded CU size
|
* until the coded block size has been achived. Calls luma and chroma filtering
|
||||||
|
* functions for each coded CU size.
|
||||||
*/
|
*/
|
||||||
void filter_deblock_cu(encoder_control* encoder, int32_t xCtb, int32_t yCtb, int8_t depth, int32_t edge)
|
void filter_deblock_cu(encoder_control *encoder, int32_t x, int32_t y, int8_t depth, int32_t edge)
|
||||||
{
|
{
|
||||||
CU_info *cur_CU = &encoder->in.cur_pic->CU[depth][xCtb+yCtb*(encoder->in.width_in_lcu<<MAX_DEPTH)];
|
CU_info *cur_cu = &encoder->in.cur_pic->CU[depth][x + y*(encoder->in.width_in_lcu << MAX_DEPTH)];
|
||||||
uint8_t split_flag = (cur_CU->depth > depth)?1:0;
|
uint8_t split_flag = (cur_cu->depth > depth) ? 1 : 0;
|
||||||
uint8_t border_x = ((encoder->in.width)<( xCtb*(LCU_WIDTH>>MAX_DEPTH) + (LCU_WIDTH>>depth) ))?1:0;
|
uint8_t border_x = (encoder->in.width < x*(LCU_WIDTH >> MAX_DEPTH) + (LCU_WIDTH >> depth)) ? 1 : 0;
|
||||||
uint8_t border_y = ((encoder->in.height)<( yCtb*(LCU_WIDTH>>MAX_DEPTH) + (LCU_WIDTH>>depth) ))?1:0;
|
uint8_t border_y = (encoder->in.height < y*(LCU_WIDTH >> MAX_DEPTH) + (LCU_WIDTH >> depth)) ? 1 : 0;
|
||||||
uint8_t border_split_x = ((encoder->in.width) < ( (xCtb+1)*(LCU_WIDTH>>MAX_DEPTH) + (LCU_WIDTH>>(depth+1)) ))?0:1;
|
uint8_t border_split_x = (encoder->in.width < ((x + 1) * (LCU_WIDTH >> MAX_DEPTH)) + (LCU_WIDTH >> (depth + 1))) ? 0 : 1;
|
||||||
uint8_t border_split_y = ((encoder->in.height) < ( (yCtb+1)*(LCU_WIDTH>>MAX_DEPTH) + (LCU_WIDTH>>(depth+1)) ))?0:1;
|
uint8_t border_split_y = (encoder->in.height < ((y + 1) * (LCU_WIDTH >> MAX_DEPTH)) + (LCU_WIDTH >> (depth + 1))) ? 0 : 1;
|
||||||
|
|
||||||
uint8_t border = border_x | border_y; /*!< are we in any border CU */
|
uint8_t border = border_x | border_y; // are we in any border CU?
|
||||||
|
|
||||||
/* split 64x64, on split flag and on border */
|
// split 64x64, on split flag and on border
|
||||||
if(!depth || split_flag || border)
|
if (depth == 0 || split_flag || border) {
|
||||||
{
|
// Split the four sub-blocks of this block recursively.
|
||||||
/* Split blocks and remember to change x and y block positions */
|
|
||||||
uint8_t change = 1 << (MAX_DEPTH - 1 - depth);
|
uint8_t change = 1 << (MAX_DEPTH - 1 - depth);
|
||||||
filter_deblock_cu(encoder,xCtb,yCtb,depth+1,edge); /* x,y */
|
filter_deblock_cu(encoder, x, y, depth + 1, edge);
|
||||||
|
if(!border_x || border_split_x) {
|
||||||
if(!border_x || border_split_x)
|
filter_deblock_cu(encoder, x + change, y, depth + 1, edge);
|
||||||
{
|
|
||||||
filter_deblock_cu(encoder,xCtb+change,yCtb,depth+1,edge); /* x+1,y */
|
|
||||||
}
|
}
|
||||||
if(!border_y || border_split_y)
|
if(!border_y || border_split_y) {
|
||||||
{
|
filter_deblock_cu(encoder, x , y + change, depth + 1, edge);
|
||||||
filter_deblock_cu(encoder,xCtb,yCtb+change,depth+1,edge); /* x,y+1 */
|
|
||||||
}
|
}
|
||||||
if((!border_x && !border_y) || (border_split_x && border_split_y) )
|
if((!border_x && !border_y) || (border_split_x && border_split_y)) {
|
||||||
{
|
filter_deblock_cu(encoder, x + change, y + change, depth + 1, edge);
|
||||||
filter_deblock_cu(encoder,xCtb+change,yCtb+change,depth+1,edge); /* x+1,y+1 */
|
|
||||||
}
|
}
|
||||||
return;
|
return;
|
||||||
}
|
}
|
||||||
/* no filtering on borders (where filter would use pixels outside the picture) */
|
|
||||||
if((!xCtb && edge == EDGE_VER) || (!yCtb && edge == EDGE_HOR)) return;
|
|
||||||
|
|
||||||
/* do the filtering for block edge */
|
// no filtering on borders (where filter would use pixels outside the picture)
|
||||||
filter_deblock_edge_luma(encoder, xCtb*(LCU_WIDTH>>MAX_DEPTH), yCtb*(LCU_WIDTH>>MAX_DEPTH), depth, edge);
|
if ((x == 0 && edge == EDGE_VER) || (y == 0 && edge == EDGE_HOR)) return;
|
||||||
filter_deblock_edge_chroma(encoder, xCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), yCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), depth, edge);
|
|
||||||
|
// do the filtering for block edge
|
||||||
|
filter_deblock_edge_luma(encoder, x*(LCU_WIDTH >> MAX_DEPTH), y*(LCU_WIDTH >> MAX_DEPTH), depth, edge);
|
||||||
|
filter_deblock_edge_chroma(encoder, x*(LCU_WIDTH >> (MAX_DEPTH + 1)), y*(LCU_WIDTH >> (MAX_DEPTH + 1)), depth, edge);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
/**
|
||||||
|
* \brief Main function for Deblocking filtering
|
||||||
/*! \brief Main function for Deblocking filtering
|
* \param encoder the encoder info structure
|
||||||
\param encoder the encoder info structure
|
*
|
||||||
|
* This is the main function for deblocking filter, it will loop through all
|
||||||
This is the main function for deblocking filter, it will loop through all the
|
* the Largest Coding Units (LCU) and call filter_deblock_cu with absolute
|
||||||
Largest Coding Units and call filter_deblock_CU with absolute X and Y coordinate
|
* X and Y coordinates of the LCU.
|
||||||
of the LCU.
|
|
||||||
*/
|
*/
|
||||||
void filter_deblock(encoder_control* encoder)
|
void filter_deblock(encoder_control* encoder)
|
||||||
{
|
{
|
||||||
int16_t xCtb,yCtb;
|
int16_t x, y;
|
||||||
|
|
||||||
/* TODO: Optimization: add thread for each LCU */
|
// TODO: Optimization: add thread for each LCU
|
||||||
/* Loop through every LCU in the slice */
|
// Filter vertically.
|
||||||
for(yCtb = 0; yCtb < encoder->in.height_in_lcu; yCtb++)
|
for (y = 0; y < encoder->in.height_in_lcu; y++)
|
||||||
{
|
{
|
||||||
for(xCtb = 0; xCtb < encoder->in.width_in_lcu; xCtb++)
|
for (x = 0; x < encoder->in.width_in_lcu; x++)
|
||||||
{
|
{
|
||||||
filter_deblock_cu(encoder, xCtb<<MAX_DEPTH, yCtb<<MAX_DEPTH, 0, EDGE_VER);
|
filter_deblock_cu(encoder, x << MAX_DEPTH, y << MAX_DEPTH, 0, EDGE_VER);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
/* Loop through every LCU in the slice */
|
// Filter horizontally.
|
||||||
for(yCtb = 0; yCtb < encoder->in.height_in_lcu; yCtb++)
|
for (y = 0; y < encoder->in.height_in_lcu; y++)
|
||||||
{
|
{
|
||||||
for(xCtb = 0; xCtb < encoder->in.width_in_lcu; xCtb++)
|
for (x = 0; x < encoder->in.width_in_lcu; x++)
|
||||||
{
|
{
|
||||||
filter_deblock_cu(encoder, xCtb<<MAX_DEPTH, yCtb<<MAX_DEPTH, 0, EDGE_HOR);
|
filter_deblock_cu(encoder, x << MAX_DEPTH, y << MAX_DEPTH, 0, EDGE_HOR);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
}
|
}
|
||||||
32
src/filter.h
32
src/filter.h
|
|
@ -17,17 +17,29 @@
|
||||||
#include "encoder.h"
|
#include "encoder.h"
|
||||||
|
|
||||||
|
|
||||||
|
//////////////////////////////////////////////////////////////////////////
|
||||||
|
// FUNCTIONS
|
||||||
|
// Deblocking
|
||||||
|
void filter_deblock_cu(encoder_control *encoder, int32_t x_cu, int32_t y_cu,
|
||||||
|
int8_t depth, int32_t edge);
|
||||||
|
void filter_deblock_edge_luma(encoder_control *encoder,
|
||||||
|
int32_t x_pos, int32_t y_pos,
|
||||||
|
int8_t depth, int8_t dir);
|
||||||
|
void filter_deblock_edge_chroma(encoder_control *encoder,
|
||||||
|
int32_t xpos, int32_t ypos,
|
||||||
|
int8_t depth, int8_t dir);
|
||||||
|
void filter_deblock(encoder_control *encoder);
|
||||||
|
void filter_deblock_luma(uint8_t *src, int32_t offset, int32_t tc , int8_t sw,
|
||||||
|
int8_t part_p_nofilter, int8_t part_q_nofilter,
|
||||||
|
int32_t thr_cut,
|
||||||
|
int8_t filter_second_p, int8_t filter_second_q);
|
||||||
|
void filter_deblock_chroma(uint8_t *src, int32_t offset, int32_t tc,
|
||||||
|
int8_t part_p_nofilter, int8_t part_q_nofilter);
|
||||||
|
// SAO
|
||||||
|
|
||||||
|
//////////////////////////////////////////////////////////////////////////
|
||||||
|
// MACROS
|
||||||
#define EDGE_VER 0
|
#define EDGE_VER 0
|
||||||
#define EDGE_HOR 1
|
#define EDGE_HOR 1
|
||||||
|
|
||||||
/* DEBLOCKING */
|
|
||||||
void filter_deblock_cu(encoder_control* encoder, int32_t x_cu, int32_t y_cu, int8_t depth, int32_t edge);
|
|
||||||
void filter_deblock_edge_luma(encoder_control* encoder, int32_t x_pos, int32_t y_pos, int8_t depth, int8_t dir);
|
|
||||||
void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t ypos, int8_t depth, int8_t dir);
|
|
||||||
void filter_deblock(encoder_control* encoder);
|
|
||||||
void filter_deblock_luma( uint8_t* src, int32_t offset, int32_t tc , int8_t sw, int8_t part_p_nofilter, int8_t part_q_nofilter, int32_t thr_cut, int8_t filter_second_p, int8_t filter_second_q);
|
|
||||||
void filter_deblock_chroma( uint8_t* src, int32_t offset, int32_t tc ,int8_t part_p_nofilter, int8_t part_q_nofilter);
|
|
||||||
|
|
||||||
/* SAO */
|
|
||||||
|
|
||||||
#endif
|
#endif
|
||||||
|
|
|
||||||
Loading…
Reference in a new issue