mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2024-11-25 11:40:44 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #6373 from terfendail:vt/stereobm_valgrind

Browse Source
This commit is contained in:
Alexander Alekhin 2016-04-13 13:38:53 +00:00
commit 5b73f3a316
2 changed files with 40 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
modules/calib3d/src/stereobm.cpp
View File

@ -344,7 +344,7 @@ static void findStereoCorrespondenceBM_SSE2( const Mat& left, const Mat& right,
{
hsad = hsad0 - dy0*ndisp; cbuf = cbuf0 + (x + wsz2 + 1)*cstep - dy0*ndisp;
lptr = lptr0 + MIN(MAX(x, -lofs), width-lofs-1) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x, -rofs), width-rofs-1) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x, -rofs), width-rofs-ndisp) - dy0*sstep;
for( y = -dy0; y < height + dy1; y++, hsad += ndisp, cbuf += ndisp, lptr += sstep, rptr += sstep )
{
@ -385,7 +385,7 @@ static void findStereoCorrespondenceBM_SSE2( const Mat& left, const Mat& right,
hsad = hsad0 - dy0*ndisp;
lptr_sub = lptr0 + MIN(MAX(x0, -lofs), width-1-lofs) - dy0*sstep;
lptr = lptr0 + MIN(MAX(x1, -lofs), width-1-lofs) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x1, -rofs), width-1-rofs) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x1, -rofs), width-ndisp-rofs) - dy0*sstep;
for( y = -dy0; y < height + dy1; y++, cbuf += ndisp, cbuf_sub += ndisp,
hsad += ndisp, lptr += sstep, lptr_sub += sstep, rptr += sstep )
@ -610,7 +610,7 @@ findStereoCorrespondenceBM( const Mat& left, const Mat& right,
{
hsad = hsad0 - dy0*ndisp; cbuf = cbuf0 + (x + wsz2 + 1)*cstep - dy0*ndisp;
lptr = lptr0 + std::min(std::max(x, -lofs), width-lofs-1) - dy0*sstep;
rptr = rptr0 + std::min(std::max(x, -rofs), width-rofs-1) - dy0*sstep;
rptr = rptr0 + std::min(std::max(x, -rofs), width-rofs-ndisp) - dy0*sstep;
for( y = -dy0; y < height + dy1; y++, hsad += ndisp, cbuf += ndisp, lptr += sstep, rptr += sstep )
{
@ -661,7 +661,7 @@ findStereoCorrespondenceBM( const Mat& left, const Mat& right,
hsad = hsad0 - dy0*ndisp;
lptr_sub = lptr0 + MIN(MAX(x0, -lofs), width-1-lofs) - dy0*sstep;
lptr = lptr0 + MIN(MAX(x1, -lofs), width-1-lofs) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x1, -rofs), width-1-rofs) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x1, -rofs), width-ndisp-rofs) - dy0*sstep;
for( y = -dy0; y < height + dy1; y++, cbuf += ndisp, cbuf_sub += ndisp,
hsad += ndisp, lptr += sstep, lptr_sub += sstep, rptr += sstep )

37
modules/calib3d/src/stereosgbm.cpp
View File

@ -99,7 +99,7 @@ StereoSGBM::~StereoSGBM()
}
/*
For each pixel row1[x], max(-maxD, 0) <= minX <= x < maxX <= width - max(0, -minD),
For each pixel row1[x], max(-maxD, 0) <= minX <= x < maxX <= width - max(0, minD),
and for each disparity minD<=d<maxD the function
computes the cost (cost[(x-minX)*(maxD - minD) + (d - minD)]), depending on the difference between
row1[x] and row2[x-d]. The subpixel algorithm from
@ -114,8 +114,8 @@ static void calcPixelCostBT( const Mat& img1, const Mat& img2, int y,
int tabOfs, int )
{
int x, c, width = img1.cols, cn = img1.channels();
int minX1 = max(-maxD, 0), maxX1 = width + min(minD, 0);
int minX2 = max(minX1 - maxD, 0), maxX2 = min(maxX1 - minD, width);
int minX1 = max(-maxD, 0), maxX1 = width + min(-minD, 0);
int minX2 = max(minX1 + minD, 0), maxX2 = min(maxX1 + maxD, width);
int D = maxD - minD, width1 = maxX1 - minX1, width2 = maxX2 - minX2;
const PixType *row1 = img1.ptr<PixType>(y), *row2 = img2.ptr<PixType>(y);
PixType *prow1 = buffer + width2*2, *prow2 = prow1 + width*cn*2;
@ -200,6 +200,19 @@ static void calcPixelCostBT( const Mat& img1, const Mat& img2, int y,
int u0 = min(ul, ur); u0 = min(u0, u);
int u1 = max(ul, ur); u1 = max(u1, u);
int minDlocal = max(minD, x-width+1);
int maxDlocal = min(maxD, x);
int d;
for( d = minD; d < minDlocal; d++ )
{
int v = prow2[0];
int v0 = buffer[0];
int v1 = buffer[width2];
int c0 = max(0, u - v1); c0 = max(c0, v0 - u);
int c1 = max(0, v - u1); c1 = max(c1, u0 - v);
cost[x*D + d] = (CostType)(cost[x*D+d] + (min(c0, c1) >> diff_scale));
}
#if CV_SSE2
if( useSIMD )
{
@ -207,7 +220,7 @@ static void calcPixelCostBT( const Mat& img1, const Mat& img2, int y,
__m128i _u1 = _mm_set1_epi8((char)u1), z = _mm_setzero_si128();
__m128i ds = _mm_cvtsi32_si128(diff_scale);
for( int d = minD; d < maxD; d += 16 )
for( ; d < maxDlocal - 15; d += 16 )
{
__m128i _v = _mm_loadu_si128((const __m128i*)(prow2 + width-x-1 + d));
__m128i _v0 = _mm_loadu_si128((const __m128i*)(buffer + width-x-1 + d));
@ -223,10 +236,8 @@ static void calcPixelCostBT( const Mat& img1, const Mat& img2, int y,
_mm_store_si128((__m128i*)(cost + x*D + d + 8), _mm_adds_epi16(c1, _mm_srl_epi16(_mm_unpackhi_epi8(diff,z), ds)));
}
}
else
#endif
{
for( int d = minD; d < maxD; d++ )
for( ; d < maxDlocal; d++ )
{
int v = prow2[width-x-1 + d];
int v0 = buffer[width-x-1 + d];
@ -236,6 +247,15 @@ static void calcPixelCostBT( const Mat& img1, const Mat& img2, int y,
cost[x*D + d] = (CostType)(cost[x*D+d] + (min(c0, c1) >> diff_scale));
}
for( ; d < maxD; d++ )
{
int v = prow2[width-1];
int v0 = buffer[width-1];
int v1 = buffer[width-1 + width2];
int c0 = max(0, u - v1); c0 = max(c0, v0 - u);
int c1 = max(0, v - u1); c1 = max(c1, u0 - v);
cost[x*D + d] = (CostType)(cost[x*D+d] + (min(c0, c1) >> diff_scale));
}
}
}
@ -329,7 +349,7 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
int disp12MaxDiff = params.disp12MaxDiff > 0 ? params.disp12MaxDiff : 1;
int P1 = params.P1 > 0 ? params.P1 : 2, P2 = max(params.P2 > 0 ? params.P2 : 5, P1+1);
int k, width = disp1.cols, height = disp1.rows;
int minX1 = max(-maxD, 0), maxX1 = width + min(minD, 0);
int minX1 = max(-maxD, 0), maxX1 = width + min(-minD, 0);
int D = maxD - minD, width1 = maxX1 - minX1;
int INVALID_DISP = minD - 1, INVALID_DISP_SCALED = INVALID_DISP*DISP_SCALE;
int SW2 = SADWindowSize.width/2, SH2 = SADWindowSize.height/2;
@ -377,6 +397,7 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
// summary cost over different (nDirs) directions
CostType* Cbuf = (CostType*)alignPtr(buffer.data, ALIGN);
memset(Cbuf, 0, CSBufSize*sizeof(CostType));
CostType* Sbuf = Cbuf + CSBufSize;
CostType* hsumBuf = Sbuf + CSBufSize;
CostType* pixDiff = hsumBuf + costBufSize*hsumBufNRows;