void cv::Sobel( InputArray _src, OutputArray _dst, int ddepth, int dx, int dy,
int ksize, double scale, double delta, int borderType )
{
int stype = _src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype);
if (ddepth < 0)
ddepth = sdepth;
int dtype = CV_MAKE_TYPE(ddepth, cn);
_dst.create( _src.size(), dtype );
#ifdef HAVE_TEGRA_OPTIMIZATION
if (tegra::useTegra() && scale == 1.0 && delta == 0)
{
Mat src = _src.getMat(), dst = _dst.getMat();
if (ksize == 3 && tegra::sobel3x3(src, dst, dx, dy, borderType))
return;
if (ksize == -1 && tegra::scharr(src, dst, dx, dy, borderType))
return;
}
#endif
#ifdef HAVE_IPP
CV_IPP_CHECK()
{
if (ksize < 0)
{
if (IPPDerivScharr(_src, _dst, ddepth, dx, dy, scale, delta, borderType))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
}
else if (0 < ksize)
{
if (IPPDerivSobel(_src, _dst, ddepth, dx, dy, ksize, scale, delta, borderType))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
}
}
#endif
int ktype = std::max(CV_32F, std::max(ddepth, sdepth));
Mat kx, ky;
getDerivKernels( kx, ky, dx, dy, ksize, false, ktype );
if( scale != 1 )
{
// usually the smoothing part is the slowest to compute,
// so try to scale it instead of the faster differenciating part
if( dx == 0 )
kx *= scale;
else
ky *= scale;
}
sepFilter2D( _src, _dst, ddepth, kx, ky, Point(-1, -1), delta, borderType );
}
Rect CylindricalWarper::buildMaps(Size src_size, InputArray K, InputArray R, OutputArray xmap, OutputArray ymap)
{
if (ocl::useOpenCL())
{
ocl::Kernel k("buildWarpCylindricalMaps", ocl::stitching::warpers_oclsrc);
if (!k.empty())
{
int rowsPerWI = ocl::Device::getDefault().isIntel() ? 4 : 1;
projector_.setCameraParams(K, R);
Point dst_tl, dst_br;
detectResultRoi(src_size, dst_tl, dst_br);
Size dsize(dst_br.x - dst_tl.x + 1, dst_br.y - dst_tl.y + 1);
xmap.create(dsize, CV_32FC1);
ymap.create(dsize, CV_32FC1);
Mat k_rinv(1, 9, CV_32FC1, projector_.k_rinv);
UMat uxmap = xmap.getUMat(), uymap = ymap.getUMat(), uk_rinv = k_rinv.getUMat(ACCESS_READ);
k.args(ocl::KernelArg::WriteOnlyNoSize(uxmap), ocl::KernelArg::WriteOnly(uymap),
ocl::KernelArg::PtrReadOnly(uk_rinv), dst_tl.x, dst_tl.y, projector_.scale,
rowsPerWI);
size_t globalsize[2] = { dsize.width, (dsize.height + rowsPerWI - 1) / rowsPerWI };
if (k.run(2, globalsize, NULL, true))
{
CV_IMPL_ADD(CV_IMPL_OCL);
return Rect(dst_tl, dst_br);
}
}
}
return RotationWarperBase<CylindricalProjector>::buildMaps(src_size, K, R, xmap, ymap);
}
void UMat::convertTo(OutputArray _dst, int _type, double alpha, double beta) const
{
bool noScale = std::fabs(alpha - 1) < DBL_EPSILON && std::fabs(beta) < DBL_EPSILON;
int stype = type(), cn = CV_MAT_CN(stype);
if( _type < 0 )
_type = _dst.fixedType() ? _dst.type() : stype;
else
_type = CV_MAKETYPE(CV_MAT_DEPTH(_type), cn);
int sdepth = CV_MAT_DEPTH(stype), ddepth = CV_MAT_DEPTH(_type);
if( sdepth == ddepth && noScale )
{
copyTo(_dst);
return;
}
#ifdef HAVE_OPENCL
bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0;
bool needDouble = sdepth == CV_64F || ddepth == CV_64F;
if( dims <= 2 && cn && _dst.isUMat() && ocl::useOpenCL() &&
((needDouble && doubleSupport) || !needDouble) )
{
int wdepth = std::max(CV_32F, sdepth), rowsPerWI = 4;
char cvt[2][40];
ocl::Kernel k("convertTo", ocl::core::convert_oclsrc,
format("-D srcT=%s -D WT=%s -D dstT=%s -D convertToWT=%s -D convertToDT=%s%s%s",
ocl::typeToStr(sdepth), ocl::typeToStr(wdepth), ocl::typeToStr(ddepth),
ocl::convertTypeStr(sdepth, wdepth, 1, cvt[0]),
ocl::convertTypeStr(wdepth, ddepth, 1, cvt[1]),
doubleSupport ? " -D DOUBLE_SUPPORT" : "", noScale ? " -D NO_SCALE" : ""));
if (!k.empty())
{
UMat src = *this;
_dst.create( size(), _type );
UMat dst = _dst.getUMat();
float alphaf = (float)alpha, betaf = (float)beta;
ocl::KernelArg srcarg = ocl::KernelArg::ReadOnlyNoSize(src),
dstarg = ocl::KernelArg::WriteOnly(dst, cn);
if (noScale)
k.args(srcarg, dstarg, rowsPerWI);
else if (wdepth == CV_32F)
k.args(srcarg, dstarg, alphaf, betaf, rowsPerWI);
else
k.args(srcarg, dstarg, alpha, beta, rowsPerWI);
size_t globalsize[2] = { (size_t)dst.cols * cn, ((size_t)dst.rows + rowsPerWI - 1) / rowsPerWI };
if (k.run(2, globalsize, NULL, false))
{
CV_IMPL_ADD(CV_IMPL_OCL);
return;
}
}
}
#endif
Mat m = getMat(ACCESS_READ);
m.convertTo(_dst, _type, alpha, beta);
}
CV_IPP_CHECK()
{
if (ippiCopy_8u_C1MR(_src, (int)sstep, _dst, (int)dstep, ippiSize(size), mask, (int)mstep) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
}
UMat& UMat::setTo(InputArray _value, InputArray _mask)
{
bool haveMask = !_mask.empty();
#ifdef HAVE_OPENCL
int tp = type(), cn = CV_MAT_CN(tp), d = CV_MAT_DEPTH(tp);
if( dims <= 2 && cn <= 4 && CV_MAT_DEPTH(tp) < CV_64F && ocl::useOpenCL() )
{
Mat value = _value.getMat();
CV_Assert( checkScalar(value, type(), _value.kind(), _InputArray::UMAT) );
int kercn = haveMask || cn == 3 ? cn : std::max(cn, ocl::predictOptimalVectorWidth(*this)),
kertp = CV_MAKE_TYPE(d, kercn);
double buf[16] = { 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0 };
convertAndUnrollScalar(value, tp, (uchar *)buf, kercn / cn);
int scalarcn = kercn == 3 ? 4 : kercn, rowsPerWI = ocl::Device::getDefault().isIntel() ? 4 : 1;
String opts = format("-D dstT=%s -D rowsPerWI=%d -D dstST=%s -D dstT1=%s -D cn=%d",
ocl::memopTypeToStr(kertp), rowsPerWI,
ocl::memopTypeToStr(CV_MAKETYPE(d, scalarcn)),
ocl::memopTypeToStr(d), kercn);
ocl::Kernel setK(haveMask ? "setMask" : "set", ocl::core::copyset_oclsrc, opts);
if( !setK.empty() )
{
ocl::KernelArg scalararg(0, 0, 0, 0, buf, CV_ELEM_SIZE(d) * scalarcn);
UMat mask;
if( haveMask )
{
mask = _mask.getUMat();
CV_Assert( mask.size() == size() && mask.type() == CV_8UC1 );
ocl::KernelArg maskarg = ocl::KernelArg::ReadOnlyNoSize(mask),
dstarg = ocl::KernelArg::ReadWrite(*this);
setK.args(maskarg, dstarg, scalararg);
}
else
{
ocl::KernelArg dstarg = ocl::KernelArg::WriteOnly(*this, cn, kercn);
setK.args(dstarg, scalararg);
}
size_t globalsize[] = { cols * cn / kercn, (rows + rowsPerWI - 1) / rowsPerWI };
if( setK.run(2, globalsize, NULL, false) )
{
CV_IMPL_ADD(CV_IMPL_OCL);
return *this;
}
}
}
#endif
Mat m = getMat(haveMask ? ACCESS_RW : ACCESS_WRITE);
m.setTo(_value, _mask);
return *this;
}
inline int arithm_ipp_not8u(const uchar* src1, size_t step1, uchar* dst, size_t step, int width, int height)
{
if (!CV_IPP_CHECK_COND)
return 0;
if (height == 1)
step1 = step = width * sizeof(dst[0]);
if (0 <= CV_INSTRUMENT_FUN_IPP(ippiNot_8u_C1R, src1, (int)step1, dst, (int)step, ippiSize(width, height)))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return 1;
}
setIppErrorStatus();
return 0;
}
Rect PlaneWarper::buildMaps(Size src_size, InputArray K, InputArray R, InputArray T, OutputArray _xmap, OutputArray _ymap)
{
projector_.setCameraParams(K, R, T);
Point dst_tl, dst_br;
detectResultRoi(src_size, dst_tl, dst_br);
Size dsize(dst_br.x - dst_tl.x + 1, dst_br.y - dst_tl.y + 1);
_xmap.create(dsize, CV_32FC1);
_ymap.create(dsize, CV_32FC1);
if (ocl::useOpenCL())
{
ocl::Kernel k("buildWarpPlaneMaps", ocl::stitching::warpers_oclsrc);
if (!k.empty())
{
int rowsPerWI = ocl::Device::getDefault().isIntel() ? 4 : 1;
Mat k_rinv(1, 9, CV_32FC1, projector_.k_rinv), t(1, 3, CV_32FC1, projector_.t);
UMat uxmap = _xmap.getUMat(), uymap = _ymap.getUMat(),
uk_rinv = k_rinv.getUMat(ACCESS_READ), ut = t.getUMat(ACCESS_READ);
k.args(ocl::KernelArg::WriteOnlyNoSize(uxmap), ocl::KernelArg::WriteOnly(uymap),
ocl::KernelArg::PtrReadOnly(uk_rinv), ocl::KernelArg::PtrReadOnly(ut),
dst_tl.x, dst_tl.y, projector_.scale, rowsPerWI);
size_t globalsize[2] = { dsize.width, (dsize.height + rowsPerWI - 1) / rowsPerWI };
if (k.run(2, globalsize, NULL, true))
{
CV_IMPL_ADD(CV_IMPL_OCL);
return Rect(dst_tl, dst_br);
}
}
}
Mat xmap = _xmap.getMat(), ymap = _ymap.getMat();
float x, y;
for (int v = dst_tl.y; v <= dst_br.y; ++v)
{
for (int u = dst_tl.x; u <= dst_br.x; ++u)
{
projector_.mapBackward(static_cast<float>(u), static_cast<float>(v), x, y);
xmap.at<float>(v - dst_tl.y, u - dst_tl.x) = x;
ymap.at<float>(v - dst_tl.y, u - dst_tl.x) = y;
}
}
return Rect(dst_tl, dst_br);
}
void UMat::copyTo(OutputArray _dst, InputArray _mask) const
{
if( _mask.empty() )
{
copyTo(_dst);
return;
}
#ifdef HAVE_OPENCL
int cn = channels(), mtype = _mask.type(), mdepth = CV_MAT_DEPTH(mtype), mcn = CV_MAT_CN(mtype);
CV_Assert( mdepth == CV_8U && (mcn == 1 || mcn == cn) );
if (ocl::useOpenCL() && _dst.isUMat() && dims <= 2)
{
UMatData * prevu = _dst.getUMat().u;
_dst.create( dims, size, type() );
UMat dst = _dst.getUMat();
bool haveDstUninit = false;
if( prevu != dst.u ) // do not leave dst uninitialized
haveDstUninit = true;
String opts = format("-D COPY_TO_MASK -D T1=%s -D scn=%d -D mcn=%d%s",
ocl::memopTypeToStr(depth()), cn, mcn,
haveDstUninit ? " -D HAVE_DST_UNINIT" : "");
ocl::Kernel k("copyToMask", ocl::core::copyset_oclsrc, opts);
if (!k.empty())
{
k.args(ocl::KernelArg::ReadOnlyNoSize(*this),
ocl::KernelArg::ReadOnlyNoSize(_mask.getUMat()),
haveDstUninit ? ocl::KernelArg::WriteOnly(dst) :
ocl::KernelArg::ReadWrite(dst));
size_t globalsize[2] = { cols, rows };
if (k.run(2, globalsize, NULL, false))
{
CV_IMPL_ADD(CV_IMPL_OCL);
return;
}
}
}
#endif
Mat src = getMat(ACCESS_READ);
src.copyTo(_dst, _mask);
}
static void
HoughLinesProbabilistic( Mat& image,
float rho, float theta, int threshold,
int lineLength, int lineGap,
std::vector<Vec4i>& lines, int linesMax )
{
Point pt;
float irho = 1 / rho;
RNG rng((uint64)-1);
CV_Assert( image.type() == CV_8UC1 );
int width = image.cols;
int height = image.rows;
int numangle = cvRound(CV_PI / theta);
int numrho = cvRound(((width + height) * 2 + 1) / rho);
#if defined HAVE_IPP && !defined(HAVE_IPP_ICV_ONLY) && IPP_VERSION_X100 >= 810 && IPP_DISABLE_BLOCK
CV_IPP_CHECK()
{
IppiSize srcSize = { width, height };
IppPointPolar delta = { rho, theta };
IppiHoughProbSpec* pSpec;
int bufferSize, specSize;
int ipp_linesMax = std::min(linesMax, numangle*numrho);
int linesCount = 0;
lines.resize(ipp_linesMax);
IppStatus ok = ippiHoughProbLineGetSize_8u_C1R(srcSize, delta, &specSize, &bufferSize);
Ipp8u* buffer = ippsMalloc_8u(bufferSize);
pSpec = (IppiHoughProbSpec*) malloc(specSize);
if (ok >= 0) ok = ippiHoughProbLineInit_8u32f_C1R(srcSize, delta, ippAlgHintNone, pSpec);
if (ok >= 0) ok = ippiHoughProbLine_8u32f_C1R(image.data, image.step, srcSize, threshold, lineLength, lineGap, (IppiPoint*) &lines[0], ipp_linesMax, &linesCount, buffer, pSpec);
free(pSpec);
ippsFree(buffer);
if (ok >= 0)
{
lines.resize(linesCount);
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
lines.clear();
setIppErrorStatus();
}
#endif
Mat accum = Mat::zeros( numangle, numrho, CV_32SC1 );
Mat mask( height, width, CV_8UC1 );
std::vector<float> trigtab(numangle*2);
for( int n = 0; n < numangle; n++ )
{
trigtab[n*2] = (float)(cos((double)n*theta) * irho);
trigtab[n*2+1] = (float)(sin((double)n*theta) * irho);
}
const float* ttab = &trigtab[0];
uchar* mdata0 = mask.ptr();
std::vector<Point> nzloc;
// stage 1. collect non-zero image points
for( pt.y = 0; pt.y < height; pt.y++ )
{
const uchar* data = image.ptr(pt.y);
uchar* mdata = mask.ptr(pt.y);
for( pt.x = 0; pt.x < width; pt.x++ )
{
if( data[pt.x] )
{
mdata[pt.x] = (uchar)1;
nzloc.push_back(pt);
}
else
mdata[pt.x] = 0;
}
}
int count = (int)nzloc.size();
// stage 2. process all the points in random order
for( ; count > 0; count-- )
{
// choose random point out of the remaining ones
int idx = rng.uniform(0, count);
int max_val = threshold-1, max_n = 0;
Point point = nzloc[idx];
Point line_end[2];
float a, b;
int* adata = accum.ptr<int>();
int i = point.y, j = point.x, k, x0, y0, dx0, dy0, xflag;
int good_line;
const int shift = 16;
// "remove" it by overriding it with the last element
nzloc[idx] = nzloc[count-1];
// check if it has been excluded already (i.e. belongs to some other line)
if( !mdata0[i*width + j] )
continue;
// update accumulator, find the most probable line
for( int n = 0; n < numangle; n++, adata += numrho )
{
int r = cvRound( j * ttab[n*2] + i * ttab[n*2+1] );
r += (numrho - 1) / 2;
int val = ++adata[r];
if( max_val < val )
{
max_val = val;
max_n = n;
}
}
// if it is too "weak" candidate, continue with another point
if( max_val < threshold )
continue;
// from the current point walk in each direction
// along the found line and extract the line segment
a = -ttab[max_n*2+1];
b = ttab[max_n*2];
x0 = j;
y0 = i;
if( fabs(a) > fabs(b) )
{
xflag = 1;
dx0 = a > 0 ? 1 : -1;
dy0 = cvRound( b*(1 << shift)/fabs(a) );
y0 = (y0 << shift) + (1 << (shift-1));
}
else
{
xflag = 0;
dy0 = b > 0 ? 1 : -1;
dx0 = cvRound( a*(1 << shift)/fabs(b) );
x0 = (x0 << shift) + (1 << (shift-1));
}
for( k = 0; k < 2; k++ )
{
int gap = 0, x = x0, y = y0, dx = dx0, dy = dy0;
if( k > 0 )
dx = -dx, dy = -dy;
// walk along the line using fixed-point arithmetics,
// stop at the image border or in case of too big gap
for( ;; x += dx, y += dy )
{
uchar* mdata;
int i1, j1;
if( xflag )
{
j1 = x;
i1 = y >> shift;
}
else
{
j1 = x >> shift;
i1 = y;
}
if( j1 < 0 || j1 >= width || i1 < 0 || i1 >= height )
break;
mdata = mdata0 + i1*width + j1;
// for each non-zero point:
// update line end,
// clear the mask element
// reset the gap
if( *mdata )
{
gap = 0;
line_end[k].y = i1;
line_end[k].x = j1;
}
else if( ++gap > lineGap )
break;
}
}
good_line = std::abs(line_end[1].x - line_end[0].x) >= lineLength ||
std::abs(line_end[1].y - line_end[0].y) >= lineLength;
for( k = 0; k < 2; k++ )
{
int x = x0, y = y0, dx = dx0, dy = dy0;
if( k > 0 )
dx = -dx, dy = -dy;
// walk along the line using fixed-point arithmetics,
// stop at the image border or in case of too big gap
for( ;; x += dx, y += dy )
{
uchar* mdata;
int i1, j1;
if( xflag )
{
j1 = x;
i1 = y >> shift;
}
else
{
j1 = x >> shift;
i1 = y;
}
mdata = mdata0 + i1*width + j1;
// for each non-zero point:
// update line end,
// clear the mask element
// reset the gap
if( *mdata )
{
if( good_line )
{
adata = accum.ptr<int>();
for( int n = 0; n < numangle; n++, adata += numrho )
{
int r = cvRound( j1 * ttab[n*2] + i1 * ttab[n*2+1] );
r += (numrho - 1) / 2;
adata[r]--;
}
}
*mdata = 0;
}
if( i1 == line_end[k].y && j1 == line_end[k].x )
break;
}
}
static double
getThreshVal_Otsu_8u( const Mat& _src )
{
Size size = _src.size();
int step = (int) _src.step;
if( _src.isContinuous() )
{
size.width *= size.height;
size.height = 1;
step = size.width;
}
#if IPP_VERSION_X100 >= 801 && !defined(HAVE_IPP_ICV_ONLY)
CV_IPP_CHECK()
{
IppiSize srcSize = { size.width, size.height };
Ipp8u thresh;
CV_SUPPRESS_DEPRECATED_START
IppStatus ok = ippiComputeThreshold_Otsu_8u_C1R(_src.ptr(), step, srcSize, &thresh);
CV_SUPPRESS_DEPRECATED_END
if (ok >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return thresh;
}
setIppErrorStatus();
}
#endif
const int N = 256;
int i, j, h[N] = {0};
for( i = 0; i < size.height; i++ )
{
const uchar* src = _src.ptr() + step*i;
j = 0;
#if CV_ENABLE_UNROLLED
for( ; j <= size.width - 4; j += 4 )
{
int v0 = src[j], v1 = src[j+1];
h[v0]++; h[v1]++;
v0 = src[j+2]; v1 = src[j+3];
h[v0]++; h[v1]++;
}
#endif
for( ; j < size.width; j++ )
h[src[j]]++;
}
double mu = 0, scale = 1./(size.width*size.height);
for( i = 0; i < N; i++ )
mu += i*(double)h[i];
mu *= scale;
double mu1 = 0, q1 = 0;
double max_sigma = 0, max_val = 0;
for( i = 0; i < N; i++ )
{
double p_i, q2, mu2, sigma;
p_i = h[i]*scale;
mu1 *= q1;
q1 += p_i;
q2 = 1. - q1;
if( std::min(q1,q2) < FLT_EPSILON || std::max(q1,q2) > 1. - FLT_EPSILON )
continue;
mu1 = (mu1 + i*p_i)/q1;
mu2 = (mu - q1*mu1)/q2;
sigma = q1*q2*(mu1 - mu2)*(mu1 - mu2);
if( sigma > max_sigma )
{
max_sigma = sigma;
max_val = i;
}
}
return max_val;
}
static void
thresh_32f( const Mat& _src, Mat& _dst, float thresh, float maxval, int type )
{
int i, j;
Size roi = _src.size();
roi.width *= _src.channels();
const float* src = _src.ptr<float>();
float* dst = _dst.ptr<float>();
size_t src_step = _src.step/sizeof(src[0]);
size_t dst_step = _dst.step/sizeof(dst[0]);
#if CV_SSE2
volatile bool useSIMD = checkHardwareSupport(CV_CPU_SSE);
#endif
if( _src.isContinuous() && _dst.isContinuous() )
{
roi.width *= roi.height;
roi.height = 1;
}
#ifdef HAVE_TEGRA_OPTIMIZATION
if (tegra::thresh_32f(_src, _dst, roi.width, roi.height, thresh, maxval, type))
return;
#endif
#if defined(HAVE_IPP)
CV_IPP_CHECK()
{
IppiSize sz = { roi.width, roi.height };
switch( type )
{
case THRESH_TRUNC:
if (0 <= ippiThreshold_GT_32f_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
case THRESH_TOZERO:
if (0 <= ippiThreshold_LTVal_32f_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh+FLT_EPSILON, 0))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
case THRESH_TOZERO_INV:
if (0 <= ippiThreshold_GTVal_32f_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh, 0))
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
}
}
#endif
switch( type )
{
case THRESH_BINARY:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128 thresh4 = _mm_set1_ps(thresh), maxval4 = _mm_set1_ps(maxval);
for( ; j <= roi.width - 8; j += 8 )
{
__m128 v0, v1;
v0 = _mm_loadu_ps( src + j );
v1 = _mm_loadu_ps( src + j + 4 );
v0 = _mm_cmpgt_ps( v0, thresh4 );
v1 = _mm_cmpgt_ps( v1, thresh4 );
v0 = _mm_and_ps( v0, maxval4 );
v1 = _mm_and_ps( v1, maxval4 );
_mm_storeu_ps( dst + j, v0 );
_mm_storeu_ps( dst + j + 4, v1 );
}
}
#elif CV_NEON
float32x4_t v_thresh = vdupq_n_f32(thresh);
uint32x4_t v_maxval = vreinterpretq_u32_f32(vdupq_n_f32(maxval));
for( ; j <= roi.width - 4; j += 4 )
{
float32x4_t v_src = vld1q_f32(src + j);
uint32x4_t v_dst = vandq_u32(vcgtq_f32(v_src, v_thresh), v_maxval);
vst1q_f32(dst + j, vreinterpretq_f32_u32(v_dst));
}
#endif
for( ; j < roi.width; j++ )
dst[j] = src[j] > thresh ? maxval : 0;
}
break;
case THRESH_BINARY_INV:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128 thresh4 = _mm_set1_ps(thresh), maxval4 = _mm_set1_ps(maxval);
for( ; j <= roi.width - 8; j += 8 )
{
__m128 v0, v1;
v0 = _mm_loadu_ps( src + j );
v1 = _mm_loadu_ps( src + j + 4 );
v0 = _mm_cmple_ps( v0, thresh4 );
v1 = _mm_cmple_ps( v1, thresh4 );
v0 = _mm_and_ps( v0, maxval4 );
v1 = _mm_and_ps( v1, maxval4 );
_mm_storeu_ps( dst + j, v0 );
_mm_storeu_ps( dst + j + 4, v1 );
}
}
#elif CV_NEON
float32x4_t v_thresh = vdupq_n_f32(thresh);
uint32x4_t v_maxval = vreinterpretq_u32_f32(vdupq_n_f32(maxval));
for( ; j <= roi.width - 4; j += 4 )
{
float32x4_t v_src = vld1q_f32(src + j);
uint32x4_t v_dst = vandq_u32(vcleq_f32(v_src, v_thresh), v_maxval);
vst1q_f32(dst + j, vreinterpretq_f32_u32(v_dst));
}
#endif
for( ; j < roi.width; j++ )
dst[j] = src[j] <= thresh ? maxval : 0;
}
break;
case THRESH_TRUNC:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128 thresh4 = _mm_set1_ps(thresh);
for( ; j <= roi.width - 8; j += 8 )
{
__m128 v0, v1;
v0 = _mm_loadu_ps( src + j );
v1 = _mm_loadu_ps( src + j + 4 );
v0 = _mm_min_ps( v0, thresh4 );
v1 = _mm_min_ps( v1, thresh4 );
_mm_storeu_ps( dst + j, v0 );
_mm_storeu_ps( dst + j + 4, v1 );
}
}
#elif CV_NEON
float32x4_t v_thresh = vdupq_n_f32(thresh);
for( ; j <= roi.width - 4; j += 4 )
vst1q_f32(dst + j, vminq_f32(vld1q_f32(src + j), v_thresh));
#endif
for( ; j < roi.width; j++ )
dst[j] = std::min(src[j], thresh);
}
break;
case THRESH_TOZERO:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128 thresh4 = _mm_set1_ps(thresh);
for( ; j <= roi.width - 8; j += 8 )
{
__m128 v0, v1;
v0 = _mm_loadu_ps( src + j );
v1 = _mm_loadu_ps( src + j + 4 );
v0 = _mm_and_ps(v0, _mm_cmpgt_ps(v0, thresh4));
v1 = _mm_and_ps(v1, _mm_cmpgt_ps(v1, thresh4));
_mm_storeu_ps( dst + j, v0 );
_mm_storeu_ps( dst + j + 4, v1 );
}
}
#elif CV_NEON
float32x4_t v_thresh = vdupq_n_f32(thresh);
for( ; j <= roi.width - 4; j += 4 )
{
float32x4_t v_src = vld1q_f32(src + j);
uint32x4_t v_dst = vandq_u32(vcgtq_f32(v_src, v_thresh),
vreinterpretq_u32_f32(v_src));
vst1q_f32(dst + j, vreinterpretq_f32_u32(v_dst));
}
#endif
for( ; j < roi.width; j++ )
{
float v = src[j];
dst[j] = v > thresh ? v : 0;
}
}
break;
case THRESH_TOZERO_INV:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128 thresh4 = _mm_set1_ps(thresh);
for( ; j <= roi.width - 8; j += 8 )
{
__m128 v0, v1;
v0 = _mm_loadu_ps( src + j );
v1 = _mm_loadu_ps( src + j + 4 );
v0 = _mm_and_ps(v0, _mm_cmple_ps(v0, thresh4));
v1 = _mm_and_ps(v1, _mm_cmple_ps(v1, thresh4));
_mm_storeu_ps( dst + j, v0 );
_mm_storeu_ps( dst + j + 4, v1 );
}
}
#elif CV_NEON
float32x4_t v_thresh = vdupq_n_f32(thresh);
for( ; j <= roi.width - 4; j += 4 )
{
float32x4_t v_src = vld1q_f32(src + j);
uint32x4_t v_dst = vandq_u32(vcleq_f32(v_src, v_thresh),
vreinterpretq_u32_f32(v_src));
vst1q_f32(dst + j, vreinterpretq_f32_u32(v_dst));
}
#endif
for( ; j < roi.width; j++ )
{
float v = src[j];
dst[j] = v <= thresh ? v : 0;
}
}
break;
default:
return CV_Error( CV_StsBadArg, "" );
}
}
static void
thresh_16s( const Mat& _src, Mat& _dst, short thresh, short maxval, int type )
{
int i, j;
Size roi = _src.size();
roi.width *= _src.channels();
const short* src = _src.ptr<short>();
short* dst = _dst.ptr<short>();
size_t src_step = _src.step/sizeof(src[0]);
size_t dst_step = _dst.step/sizeof(dst[0]);
#if CV_SSE2
volatile bool useSIMD = checkHardwareSupport(CV_CPU_SSE);
#endif
if( _src.isContinuous() && _dst.isContinuous() )
{
roi.width *= roi.height;
roi.height = 1;
src_step = dst_step = roi.width;
}
#ifdef HAVE_TEGRA_OPTIMIZATION
if (tegra::thresh_16s(_src, _dst, roi.width, roi.height, thresh, maxval, type))
return;
#endif
#if defined(HAVE_IPP)
CV_IPP_CHECK()
{
IppiSize sz = { roi.width, roi.height };
CV_SUPPRESS_DEPRECATED_START
switch( type )
{
case THRESH_TRUNC:
#ifndef HAVE_IPP_ICV_ONLY
if (_src.data == _dst.data && ippiThreshold_GT_16s_C1IR(dst, (int)dst_step*sizeof(dst[0]), sz, thresh) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
#endif
if (ippiThreshold_GT_16s_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
case THRESH_TOZERO:
#ifndef HAVE_IPP_ICV_ONLY
if (_src.data == _dst.data && ippiThreshold_LTVal_16s_C1IR(dst, (int)dst_step*sizeof(dst[0]), sz, thresh + 1, 0) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
#endif
if (ippiThreshold_LTVal_16s_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh+1, 0) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
case THRESH_TOZERO_INV:
#ifndef HAVE_IPP_ICV_ONLY
if (_src.data == _dst.data && ippiThreshold_GTVal_16s_C1IR(dst, (int)dst_step*sizeof(dst[0]), sz, thresh, 0) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
#endif
if (ippiThreshold_GTVal_16s_C1R(src, (int)src_step*sizeof(src[0]), dst, (int)dst_step*sizeof(dst[0]), sz, thresh, 0) >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
break;
}
CV_SUPPRESS_DEPRECATED_END
}
#endif
switch( type )
{
case THRESH_BINARY:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128i thresh8 = _mm_set1_epi16(thresh), maxval8 = _mm_set1_epi16(maxval);
for( ; j <= roi.width - 16; j += 16 )
{
__m128i v0, v1;
v0 = _mm_loadu_si128( (const __m128i*)(src + j) );
v1 = _mm_loadu_si128( (const __m128i*)(src + j + 8) );
v0 = _mm_cmpgt_epi16( v0, thresh8 );
v1 = _mm_cmpgt_epi16( v1, thresh8 );
v0 = _mm_and_si128( v0, maxval8 );
v1 = _mm_and_si128( v1, maxval8 );
_mm_storeu_si128((__m128i*)(dst + j), v0 );
_mm_storeu_si128((__m128i*)(dst + j + 8), v1 );
}
}
#elif CV_NEON
int16x8_t v_thresh = vdupq_n_s16(thresh), v_maxval = vdupq_n_s16(maxval);
for( ; j <= roi.width - 8; j += 8 )
{
uint16x8_t v_mask = vcgtq_s16(vld1q_s16(src + j), v_thresh);
vst1q_s16(dst + j, vandq_s16(vreinterpretq_s16_u16(v_mask), v_maxval));
}
#endif
for( ; j < roi.width; j++ )
dst[j] = src[j] > thresh ? maxval : 0;
}
break;
case THRESH_BINARY_INV:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128i thresh8 = _mm_set1_epi16(thresh), maxval8 = _mm_set1_epi16(maxval);
for( ; j <= roi.width - 16; j += 16 )
{
__m128i v0, v1;
v0 = _mm_loadu_si128( (const __m128i*)(src + j) );
v1 = _mm_loadu_si128( (const __m128i*)(src + j + 8) );
v0 = _mm_cmpgt_epi16( v0, thresh8 );
v1 = _mm_cmpgt_epi16( v1, thresh8 );
v0 = _mm_andnot_si128( v0, maxval8 );
v1 = _mm_andnot_si128( v1, maxval8 );
_mm_storeu_si128((__m128i*)(dst + j), v0 );
_mm_storeu_si128((__m128i*)(dst + j + 8), v1 );
}
}
#elif CV_NEON
int16x8_t v_thresh = vdupq_n_s16(thresh), v_maxval = vdupq_n_s16(maxval);
for( ; j <= roi.width - 8; j += 8 )
{
uint16x8_t v_mask = vcleq_s16(vld1q_s16(src + j), v_thresh);
vst1q_s16(dst + j, vandq_s16(vreinterpretq_s16_u16(v_mask), v_maxval));
}
#endif
for( ; j < roi.width; j++ )
dst[j] = src[j] <= thresh ? maxval : 0;
}
break;
case THRESH_TRUNC:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128i thresh8 = _mm_set1_epi16(thresh);
for( ; j <= roi.width - 16; j += 16 )
{
__m128i v0, v1;
v0 = _mm_loadu_si128( (const __m128i*)(src + j) );
v1 = _mm_loadu_si128( (const __m128i*)(src + j + 8) );
v0 = _mm_min_epi16( v0, thresh8 );
v1 = _mm_min_epi16( v1, thresh8 );
_mm_storeu_si128((__m128i*)(dst + j), v0 );
_mm_storeu_si128((__m128i*)(dst + j + 8), v1 );
}
}
#elif CV_NEON
int16x8_t v_thresh = vdupq_n_s16(thresh);
for( ; j <= roi.width - 8; j += 8 )
vst1q_s16(dst + j, vminq_s16(vld1q_s16(src + j), v_thresh));
#endif
for( ; j < roi.width; j++ )
dst[j] = std::min(src[j], thresh);
}
break;
case THRESH_TOZERO:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128i thresh8 = _mm_set1_epi16(thresh);
for( ; j <= roi.width - 16; j += 16 )
{
__m128i v0, v1;
v0 = _mm_loadu_si128( (const __m128i*)(src + j) );
v1 = _mm_loadu_si128( (const __m128i*)(src + j + 8) );
v0 = _mm_and_si128(v0, _mm_cmpgt_epi16(v0, thresh8));
v1 = _mm_and_si128(v1, _mm_cmpgt_epi16(v1, thresh8));
_mm_storeu_si128((__m128i*)(dst + j), v0 );
_mm_storeu_si128((__m128i*)(dst + j + 8), v1 );
}
}
#elif CV_NEON
int16x8_t v_thresh = vdupq_n_s16(thresh);
for( ; j <= roi.width - 8; j += 8 )
{
int16x8_t v_src = vld1q_s16(src + j);
uint16x8_t v_mask = vcgtq_s16(v_src, v_thresh);
vst1q_s16(dst + j, vandq_s16(vreinterpretq_s16_u16(v_mask), v_src));
}
#endif
for( ; j < roi.width; j++ )
{
short v = src[j];
dst[j] = v > thresh ? v : 0;
}
}
break;
case THRESH_TOZERO_INV:
for( i = 0; i < roi.height; i++, src += src_step, dst += dst_step )
{
j = 0;
#if CV_SSE2
if( useSIMD )
{
__m128i thresh8 = _mm_set1_epi16(thresh);
for( ; j <= roi.width - 16; j += 16 )
{
__m128i v0, v1;
v0 = _mm_loadu_si128( (const __m128i*)(src + j) );
v1 = _mm_loadu_si128( (const __m128i*)(src + j + 8) );
v0 = _mm_andnot_si128(_mm_cmpgt_epi16(v0, thresh8), v0);
v1 = _mm_andnot_si128(_mm_cmpgt_epi16(v1, thresh8), v1);
_mm_storeu_si128((__m128i*)(dst + j), v0 );
_mm_storeu_si128((__m128i*)(dst + j + 8), v1 );
}
}
#elif CV_NEON
int16x8_t v_thresh = vdupq_n_s16(thresh);
for( ; j <= roi.width - 8; j += 8 )
{
int16x8_t v_src = vld1q_s16(src + j);
uint16x8_t v_mask = vcleq_s16(v_src, v_thresh);
vst1q_s16(dst + j, vandq_s16(vreinterpretq_s16_u16(v_mask), v_src));
}
#endif
for( ; j < roi.width; j++ )
{
short v = src[j];
dst[j] = v <= thresh ? v : 0;
}
}
break;
default:
return CV_Error( CV_StsBadArg, "" );
}
}
void cv::Laplacian( InputArray _src, OutputArray _dst, int ddepth, int ksize,
double scale, double delta, int borderType )
{
int stype = _src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype);
if (ddepth < 0)
ddepth = sdepth;
_dst.create( _src.size(), CV_MAKETYPE(ddepth, cn) );
#ifdef HAVE_IPP
CV_IPP_CHECK()
{
if ((ksize == 3 || ksize == 5) && ((borderType & BORDER_ISOLATED) != 0 || !_src.isSubmatrix()) &&
((stype == CV_8UC1 && ddepth == CV_16S) || (ddepth == CV_32F && stype == CV_32FC1)) && !ocl::useOpenCL())
{
int iscale = saturate_cast<int>(scale), idelta = saturate_cast<int>(delta);
bool floatScale = std::fabs(scale - iscale) > DBL_EPSILON, needScale = iscale != 1;
bool floatDelta = std::fabs(delta - idelta) > DBL_EPSILON, needDelta = delta != 0;
int borderTypeNI = borderType & ~BORDER_ISOLATED;
Mat src = _src.getMat(), dst = _dst.getMat();
if (src.data != dst.data)
{
Ipp32s bufsize;
IppStatus status = (IppStatus)-1;
IppiSize roisize = { src.cols, src.rows };
IppiMaskSize masksize = ksize == 3 ? ippMskSize3x3 : ippMskSize5x5;
IppiBorderType borderTypeIpp = ippiGetBorderType(borderTypeNI);
#define IPP_FILTER_LAPLACIAN(ippsrctype, ippdsttype, ippfavor) \
do \
{ \
if (borderTypeIpp >= 0 && ippiFilterLaplacianGetBufferSize_##ippfavor##_C1R(roisize, masksize, &bufsize) >= 0) \
{ \
Ipp8u * buffer = ippsMalloc_8u(bufsize); \
status = ippiFilterLaplacianBorder_##ippfavor##_C1R(src.ptr<ippsrctype>(), (int)src.step, dst.ptr<ippdsttype>(), \
(int)dst.step, roisize, masksize, borderTypeIpp, 0, buffer); \
ippsFree(buffer); \
} \
} while ((void)0, 0)
CV_SUPPRESS_DEPRECATED_START
if (sdepth == CV_8U && ddepth == CV_16S && !floatScale && !floatDelta)
{
IPP_FILTER_LAPLACIAN(Ipp8u, Ipp16s, 8u16s);
if (needScale && status >= 0)
status = ippiMulC_16s_C1IRSfs((Ipp16s)iscale, dst.ptr<Ipp16s>(), (int)dst.step, roisize, 0);
if (needDelta && status >= 0)
status = ippiAddC_16s_C1IRSfs((Ipp16s)idelta, dst.ptr<Ipp16s>(), (int)dst.step, roisize, 0);
}
else if (sdepth == CV_32F && ddepth == CV_32F)
{
IPP_FILTER_LAPLACIAN(Ipp32f, Ipp32f, 32f);
if (needScale && status >= 0)
status = ippiMulC_32f_C1IR((Ipp32f)scale, dst.ptr<Ipp32f>(), (int)dst.step, roisize);
if (needDelta && status >= 0)
status = ippiAddC_32f_C1IR((Ipp32f)delta, dst.ptr<Ipp32f>(), (int)dst.step, roisize);
}
CV_SUPPRESS_DEPRECATED_END
if (status >= 0)
{
CV_IMPL_ADD(CV_IMPL_IPP);
return;
}
setIppErrorStatus();
}
}
