/*The function used to fill in the chroma plane motion vectors for a macro
block when 4 different motion vectors are specified in the luma plane.
This version is for use with chroma decimated in the X and Y directions
(4:2:0).
_cbmvs: The chroma block-level motion vectors to fill in.
_lbmvs: The luma block-level motion vectors.*/
static void oc_set_chroma_mvs00(oc_mv _cbmvs[4],const oc_mv _lbmvs[4]){
int dx;
int dy;
dx=OC_MV_X(_lbmvs[0])+OC_MV_X(_lbmvs[1])
+OC_MV_X(_lbmvs[2])+OC_MV_X(_lbmvs[3]);
dy=OC_MV_Y(_lbmvs[0])+OC_MV_Y(_lbmvs[1])
+OC_MV_Y(_lbmvs[2])+OC_MV_Y(_lbmvs[3]);
_cbmvs[0]=OC_MV(OC_DIV_ROUND_POW2(dx,2,2),OC_DIV_ROUND_POW2(dy,2,2));
}
static void oc_mcenc_find_candidates_a(oc_enc_ctx *_enc,oc_mcenc_ctx *_mcenc,
oc_mv _accum,int _mbi,int _frame){
oc_mb_enc_info *embs;
int accum_x;
int accum_y;
int a[3][2];
int ncandidates;
unsigned nmbi;
int i;
embs=_enc->mb_info;
/*Skip a position to store the median predictor in.*/
ncandidates=1;
if(embs[_mbi].ncneighbors>0){
/*Fill in the first part of set A: the vectors from adjacent blocks.*/
for(i=0;i<embs[_mbi].ncneighbors;i++){
nmbi=embs[_mbi].cneighbors[i];
_mcenc->candidates[ncandidates][0]=
OC_MV_X(embs[nmbi].analysis_mv[0][_frame]);
_mcenc->candidates[ncandidates][1]=
OC_MV_Y(embs[nmbi].analysis_mv[0][_frame]);
ncandidates++;
}
}
accum_x=OC_MV_X(_accum);
accum_y=OC_MV_Y(_accum);
/*Add a few additional vectors to set A: the vectors used in the previous
frames and the (0,0) vector.*/
_mcenc->candidates[ncandidates][0]=accum_x;
_mcenc->candidates[ncandidates][1]=accum_y;
ncandidates++;
_mcenc->candidates[ncandidates][0]=OC_CLAMPI(-31,
OC_MV_X(embs[_mbi].analysis_mv[1][_frame])+accum_x,31);
_mcenc->candidates[ncandidates][1]=OC_CLAMPI(-31,
OC_MV_Y(embs[_mbi].analysis_mv[1][_frame])+accum_y,31);
ncandidates++;
_mcenc->candidates[ncandidates][0]=0;
_mcenc->candidates[ncandidates][1]=0;
ncandidates++;
/*Use the first three vectors of set A to find our best predictor: their
median.*/
memcpy(a,_mcenc->candidates+1,sizeof(a));
OC_SORT2I(a[0][0],a[1][0]);
OC_SORT2I(a[0][1],a[1][1]);
OC_SORT2I(a[1][0],a[2][0]);
OC_SORT2I(a[1][1],a[2][1]);
OC_SORT2I(a[0][0],a[1][0]);
OC_SORT2I(a[0][1],a[1][1]);
_mcenc->candidates[0][0]=a[1][0];
_mcenc->candidates[0][1]=a[1][1];
_mcenc->setb0=ncandidates;
}
static void oc_mcenc_find_candidates_b(oc_enc_ctx *_enc,oc_mcenc_ctx *_mcenc,
oc_mv _accum,int _mbi,int _frame){
oc_mb_enc_info *embs;
int accum_x;
int accum_y;
int ncandidates;
embs=_enc->mb_info;
accum_x=OC_MV_X(_accum);
accum_y=OC_MV_Y(_accum);
/*Fill in set B: accelerated predictors for this and adjacent macro blocks.*/
ncandidates=_mcenc->setb0;
/*Use only the current block. Using more did not appear to be helpful
with the current selection logic due to escaping the local search too
quickly.*/
_mcenc->candidates[ncandidates][0]=OC_CLAMPI(-31,
2*OC_MV_X(embs[_mbi].analysis_mv[1][_frame])
-OC_MV_X(embs[_mbi].analysis_mv[2][_frame])+accum_x,31);
_mcenc->candidates[ncandidates][1]=OC_CLAMPI(-31,
2*OC_MV_Y(embs[_mbi].analysis_mv[1][_frame])
-OC_MV_Y(embs[_mbi].analysis_mv[2][_frame])+accum_y,31);
ncandidates++;
_mcenc->ncandidates=ncandidates;
}
static void oc_mcenc_find_candidates(oc_enc_ctx *_enc,oc_mcenc_ctx *_mcenc,
oc_mv _accum,int _mbi,int _frame){
oc_mb_enc_info *embs;
int accum_x;
int accum_y;
int a[3][2];
int ncandidates;
unsigned nmbi;
int i;
embs=_enc->mb_info;
/*Skip a position to store the median predictor in.*/
ncandidates=1;
if(embs[_mbi].ncneighbors>0){
/*Fill in the first part of set A: the vectors from adjacent blocks.*/
for(i=0;i<embs[_mbi].ncneighbors;i++){
nmbi=embs[_mbi].cneighbors[i];
_mcenc->candidates[ncandidates][0]=
OC_MV_X(embs[nmbi].analysis_mv[0][_frame]);
_mcenc->candidates[ncandidates][1]=
OC_MV_Y(embs[nmbi].analysis_mv[0][_frame]);
ncandidates++;
}
}
accum_x=OC_MV_X(_accum);
accum_y=OC_MV_Y(_accum);
/*Add a few additional vectors to set A: the vectors used in the previous
frames and the (0,0) vector.*/
_mcenc->candidates[ncandidates][0]=OC_CLAMPI(-31,accum_x,31);
_mcenc->candidates[ncandidates][1]=OC_CLAMPI(-31,accum_y,31);
ncandidates++;
_mcenc->candidates[ncandidates][0]=OC_CLAMPI(-31,
OC_MV_X(embs[_mbi].analysis_mv[1][_frame])+accum_x,31);
_mcenc->candidates[ncandidates][1]=OC_CLAMPI(-31,
OC_MV_Y(embs[_mbi].analysis_mv[1][_frame])+accum_y,31);
ncandidates++;
_mcenc->candidates[ncandidates][0]=0;
_mcenc->candidates[ncandidates][1]=0;
ncandidates++;
/*Use the first three vectors of set A to find our best predictor: their
median.*/
memcpy(a,_mcenc->candidates+1,sizeof(a));
OC_SORT2I(a[0][0],a[1][0]);
OC_SORT2I(a[0][1],a[1][1]);
OC_SORT2I(a[1][0],a[2][0]);
OC_SORT2I(a[1][1],a[2][1]);
OC_SORT2I(a[0][0],a[1][0]);
OC_SORT2I(a[0][1],a[1][1]);
_mcenc->candidates[0][0]=a[1][0];
_mcenc->candidates[0][1]=a[1][1];
/*Fill in set B: accelerated predictors for this and adjacent macro blocks.*/
_mcenc->setb0=ncandidates;
/*The first time through the loop use the current macro block.*/
nmbi=_mbi;
for(i=0;;i++){
_mcenc->candidates[ncandidates][0]=OC_CLAMPI(-31,
2*OC_MV_X(embs[_mbi].analysis_mv[1][_frame])
-OC_MV_X(embs[_mbi].analysis_mv[2][_frame])+accum_x,31);
_mcenc->candidates[ncandidates][1]=OC_CLAMPI(-31,
2*OC_MV_Y(embs[_mbi].analysis_mv[1][_frame])
-OC_MV_Y(embs[_mbi].analysis_mv[2][_frame])+accum_y,31);
ncandidates++;
if(i>=embs[_mbi].npneighbors)break;
nmbi=embs[_mbi].pneighbors[i];
}
/*Truncate to full-pel positions.*/
for(i=0;i<ncandidates;i++){
_mcenc->candidates[i][0]=OC_DIV2(_mcenc->candidates[i][0]);
_mcenc->candidates[i][1]=OC_DIV2(_mcenc->candidates[i][1]);
}
_mcenc->ncandidates=ncandidates;
}
