hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-27 11:24:05 +00:00
Code Issues Actions Projects Releases Wiki Activity

Fixed deblocking and more variable name cleaning

Browse source
This commit is contained in:
Marko Viitanen 2013-05-31 09:57:10 +03:00
parent d33304e5ec
commit b6daafecef
1 changed files with 60 additions and 57 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

117
src/filter.c
View file

@ -6,7 +6,7 @@
/*! \file filter.c /*! \file filter.c
\brief filtering \brief filtering
\author Marko Viitanen \author Marko Viitanen
\date 2013-04 \date 2013-05
Filtering functions Filtering functions
*/ */
@ -34,69 +34,69 @@ const uint8_t betatable_8x8[52] =
}; };
INLINE void filter_luma( uint8_t* piSrc, int32_t iOffset, int32_t tc , int8_t sw, int8_t bPartPNoFilter, int8_t bPartQNoFilter, int32_t iThrCut, int8_t bFilterSecondP, int8_t bFilterSecondQ) INLINE void filter_luma( uint8_t* src, int32_t offset, int32_t tc , int8_t sw, int8_t bPartPNoFilter, int8_t bPartQNoFilter, int32_t iThrCut, int8_t bFilterSecondP, int8_t bFilterSecondQ)
{ {
int32_t delta; int32_t delta;
int16_t m0 = piSrc[-iOffset*4]; int16_t m0 = src[-offset*4];
int16_t m1 = piSrc[-iOffset*3]; int16_t m1 = src[-offset*3];
int16_t m2 = piSrc[-iOffset*2]; int16_t m2 = src[-offset*2];
int16_t m3 = piSrc[-iOffset]; int16_t m3 = src[-offset];
int16_t m4 = piSrc[0]; int16_t m4 = src[0];
int16_t m5 = piSrc[ iOffset]; int16_t m5 = src[ offset];
int16_t m6 = piSrc[ iOffset*2]; int16_t m6 = src[ offset*2];
int16_t m7 = piSrc[ iOffset*3]; int16_t m7 = src[ offset*3];
if (sw) if (sw)
{ {
piSrc[-iOffset*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));
piSrc[-iOffset*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));
piSrc[-iOffset] = 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));
piSrc[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));
piSrc[ iOffset] = 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));
piSrc[ iOffset*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) < iThrCut ) if ( abs(delta) < iThrCut )
{ {
int32_t tc2 = tc>>1; int32_t tc2 = tc>>1;
delta = CLIP(-tc, tc, delta); delta = CLIP(-tc, tc, delta);
piSrc[-iOffset] = CLIP(0,(1 << g_bitDepth)-1,(m3+delta)); src[-offset] = CLIP(0,(1 << g_bitDepth)-1,(m3+delta));
piSrc[0] = CLIP(0,(1 << g_bitDepth)-1,(m4-delta)); src[0] = CLIP(0,(1 << g_bitDepth)-1,(m4-delta));
if(bFilterSecondP) if(bFilterSecondP)
{ {
int32_t delta1 = CLIP(-tc2, tc2, (( ((m1+m3+1)>>1)- m2+delta)>>1)); int32_t delta1 = CLIP(-tc2, tc2, (( ((m1+m3+1)>>1)- m2+delta)>>1));
piSrc[-iOffset*2] = CLIP(0,(1 << g_bitDepth)-1,(m2+delta1)); src[-offset*2] = CLIP(0,(1 << g_bitDepth)-1,(m2+delta1));
} }
if(bFilterSecondQ) if(bFilterSecondQ)
{ {
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));
piSrc[ iOffset] = CLIP(0,(1 << g_bitDepth)-1,(m5+delta2)); src[ offset] = CLIP(0,(1 << g_bitDepth)-1,(m5+delta2));
} }
} }
} }
if(bPartPNoFilter) if(bPartPNoFilter)
{ {
piSrc[-iOffset] = (uint8_t)m3; src[-offset] = (uint8_t)m3;
piSrc[-iOffset*2] = (uint8_t)m2; src[-offset*2] = (uint8_t)m2;
piSrc[-iOffset*3] = (uint8_t)m1; src[-offset*3] = (uint8_t)m1;
} }
if(bPartQNoFilter) if(bPartQNoFilter)
{ {
piSrc[0] = (uint8_t)m4; src[0] = (uint8_t)m4;
piSrc[ iOffset] = (uint8_t)m5; src[ offset] = (uint8_t)m5;
piSrc[ iOffset*2] = (uint8_t)m6; src[ offset*2] = (uint8_t)m6;
} }
} }
INLINE void filter_chroma( uint8_t* src, int32_t offset, int32_t tc ,int8_t part_p_nofilter, int8_t part_q_nofilter) INLINE void filter_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;
@ -106,11 +106,11 @@ INLINE void filter_chroma( uint8_t* src, int32_t offset, int32_t tc ,int8_t part
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);
} }
@ -118,16 +118,13 @@ INLINE void filter_chroma( uint8_t* src, int32_t offset, int32_t tc ,int8_t part
void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t ypos, int8_t depth, int8_t dir) void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t ypos, int8_t depth, int8_t dir)
{ {
int32_t iStride = encoder->in.cur_pic.width; int32_t stride = encoder->in.cur_pic.width;
int32_t offset = iStride; int32_t offset = stride;
int32_t betaOffsetDiv2 = encoder->betaOffsetdiv2; int32_t betaOffsetDiv2 = encoder->betaOffsetdiv2;
int32_t tcOffsetDiv2 = encoder->tcOffsetdiv2; int32_t tcOffsetDiv2 = encoder->tcOffsetdiv2;
const int8_t width = (LCU_WIDTH>>depth);
const int8_t scu_width = (LCU_WIDTH>>MAX_DEPTH);
const int8_t scu_width_log2 = TOBITS(scu_width);
int8_t uiBs = 2; /* Filter strength */ int8_t uiBs = 2; /* Filter strength */
/* ToDo: support 10+bits */ /* ToDo: support 10+bits */
uint8_t* origsrc = &encoder->in.cur_pic.yRecData[xpos+ypos*iStride]; uint8_t* origsrc = &encoder->in.cur_pic.yRecData[xpos+ypos*stride];
uint8_t* src = origsrc; uint8_t* src = origsrc;
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)]; //CU_info* cu = &encoder->in.cur_pic.CU[depth][(xpos>>scu_width_log2) + (ypos>>scu_width_log2)*(encoder->in.width>>scu_width_log2)];
@ -135,32 +132,31 @@ void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t yp
if(dir == EDGE_VER) if(dir == EDGE_VER)
{ {
offset = 1; offset = 1;
step = iStride; step = stride;
} }
/* For each subpart */
{ {
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*(uiBs-1) + (tcOffsetDiv2 << 1)));
int32_t B_index = CLIP(0, 51, QP + (betaOffsetDiv2 << 1)); int32_t B_index = CLIP(0, 51, QP + (betaOffsetDiv2 << 1));
int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale; int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale;
int32_t Beta = betatable_8x8[B_index]*bitdepth_scale; int32_t Beta = betatable_8x8[B_index]*bitdepth_scale;
int32_t iSideThreshold = (Beta+(Beta>>1))>>3; int32_t side_threshold = (Beta+(Beta>>1))>>3;
int32_t iThrCut = Tc*10; int32_t iThrCut = 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 (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 */
#define calc_DP(s,o) abs( s[-o*3] - 2*s[-o*2] + 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( s[0] - 2*s[o] + s[o*2] ) #define calc_DQ(s,o) abs( (int16_t)s[0] - (int16_t)2*s[o] + (int16_t)s[o*2] )
dp0 = calc_DP( (src+step*(block_idx*4+0)), offset); dp0 = calc_DP( (src+step*(block_idx*4+0)), offset);
dq0 = calc_DQ( (src+step*(block_idx*4+0)), offset); dq0 = calc_DQ( (src+step*(block_idx*4+0)), offset);
@ -177,9 +173,11 @@ void filter_deblock_edge_luma(encoder_control* encoder, int32_t xpos, int32_t yp
#endif #endif
if (d < Beta) if (d < Beta)
{ {
#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 filter_P = (dp < side_threshold)?1:0;
int8_t filter_P = (dp < iSideThreshold)?1:0; int8_t filter_Q = (dq < side_threshold)?1:0;
int8_t filter_Q = (dq < iSideThreshold)?1:0;
/* 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)) )
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)));
@ -201,7 +199,7 @@ void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t y
/* ToDo: support 10+bits */ /* ToDo: support 10+bits */
uint8_t* srcU = &encoder->in.cur_pic.uRecData[xpos+ypos*stride]; uint8_t* srcU = &encoder->in.cur_pic.uRecData[xpos+ypos*stride];
uint8_t* srcV = &encoder->in.cur_pic.vRecData[xpos+ypos*stride]; uint8_t* srcV = &encoder->in.cur_pic.vRecData[xpos+ypos*stride];
/* Init offset and step to EDGE_HOR */
int32_t offset = stride; int32_t offset = stride;
int32_t step = 1; int32_t step = 1;
@ -219,7 +217,7 @@ void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t y
// For each subpart // For each subpart
{ {
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 + (tcOffsetDiv2 << 1))); int32_t TC_index = CLIP(0, 51+2, (int32_t)(QP + 2 + (tcOffsetDiv2 << 1)));
int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale; int32_t Tc = tctable_8x8[TC_index]*bitdepth_scale;
@ -228,12 +226,12 @@ void filter_deblock_edge_chroma(encoder_control* encoder,int32_t xpos, int32_t y
for (blk_idx = 0; blk_idx < blocks_in_part; blk_idx++) for (blk_idx = 0; blk_idx < blocks_in_part; blk_idx++)
{ {
/* Chroma Red */ /* Chroma U */
filter_chroma( srcU+step*(blk_idx*4+0), offset, Tc,0, 0); filter_chroma( srcU+step*(blk_idx*4+0), offset, Tc,0, 0);
filter_chroma( srcU+step*(blk_idx*4+1), offset, Tc,0, 0); filter_chroma( srcU+step*(blk_idx*4+1), offset, Tc,0, 0);
filter_chroma( srcU+step*(blk_idx*4+2), offset, Tc,0, 0); filter_chroma( srcU+step*(blk_idx*4+2), offset, Tc,0, 0);
filter_chroma( srcU+step*(blk_idx*4+3), offset, Tc,0, 0); filter_chroma( srcU+step*(blk_idx*4+3), offset, Tc,0, 0);
/* Chroma Blue */ /* Chroma V */
filter_chroma( srcV+step*(blk_idx*4+0), offset, Tc,0, 0); filter_chroma( srcV+step*(blk_idx*4+0), offset, Tc,0, 0);
filter_chroma( srcV+step*(blk_idx*4+1), offset, Tc,0, 0); filter_chroma( srcV+step*(blk_idx*4+1), offset, Tc,0, 0);
filter_chroma( srcV+step*(blk_idx*4+2), offset, Tc,0, 0); filter_chroma( srcV+step*(blk_idx*4+2), offset, Tc,0, 0);
@ -275,12 +273,17 @@ void filter_deblock_CU(encoder_control* encoder, int32_t xCtb, int32_t yCtb, int
/* DO THE FILTERING FOR EDGE */ /* DO THE FILTERING FOR EDGE */
filter_deblock_edge_luma(encoder, xCtb*(LCU_WIDTH>>MAX_DEPTH), yCtb*(LCU_WIDTH>>MAX_DEPTH), depth, edge); filter_deblock_edge_luma(encoder, xCtb*(LCU_WIDTH>>MAX_DEPTH), yCtb*(LCU_WIDTH>>MAX_DEPTH), depth, edge);
//if ( depth != MAX_DEPTH )
{
filter_deblock_edge_chroma(encoder, xCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), yCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), depth, edge);
}
filter_deblock_edge_chroma(encoder, xCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), yCtb*(LCU_WIDTH>>(MAX_DEPTH+1)), depth, edge);
} }
/*! \brief Main function for Deblocking filtering
\param encoder the encoder info structure
This is the main function for deblocking filter, it will loop through all the
Largest Coding Units and call filter_deblock_CU with absolute X and Y coordinate
of the LCU.
*/
void filter_deblock(encoder_control* encoder) void filter_deblock(encoder_control* encoder)
{ {
int16_t xCtb,yCtb; int16_t xCtb,yCtb;