Merge release 2.4.4

This commit is contained in:
Andrey Kamaev
2013-03-04 19:29:40 +04:00
64 changed files with 11575 additions and 3169 deletions

View File

@@ -648,7 +648,7 @@ namespace cv { namespace gpu { namespace device
tWeight += gmm_weight(mode * frame.rows + y, x);
if (tWeight > c_TB)
break;
};
}
}
fgmask(y, x) = background ? 0 : isShadow ? c_shadowVal : 255;
@@ -761,4 +761,4 @@ namespace cv { namespace gpu { namespace device
}}}
#endif /* CUDA_DISABLER */
#endif /* CUDA_DISABLER */

View File

@@ -194,10 +194,10 @@ namespace cv { namespace gpu { namespace device
if ( y > 0 && connected(intensity, image(y - 1, x)))
c |= UP;
if ( x - 1 < image.cols && connected(intensity, image(y, x + 1)))
if ( x + 1 < image.cols && connected(intensity, image(y, x + 1)))
c |= RIGHT;
if ( y - 1 < image.rows && connected(intensity, image(y + 1, x)))
if ( y + 1 < image.rows && connected(intensity, image(y + 1, x)))
c |= DOWN;
components(y, x) = c;

View File

@@ -2284,15 +2284,18 @@ namespace arithm
template void bitScalarAnd<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarAnd<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarAnd<uint>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarAnd<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarAnd<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarOr<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarOr<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarOr<uint>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarOr<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarOr<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarXor<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarXor<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarXor<uint>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarXor<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
template void bitScalarXor<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
}
//////////////////////////////////////////////////////////////////////////

View File

@@ -2280,11 +2280,11 @@ namespace
{
typedef void (*bit_scalar_func_t)(PtrStepSzb src1, unsigned int src2, PtrStepSzb dst, cudaStream_t stream);
template <bit_scalar_func_t func> struct BitScalar
template <typename T, bit_scalar_func_t func> struct BitScalar
{
static void call(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream)
{
func(src, static_cast<unsigned int>(sc.val[0]), dst, stream);
func(src, saturate_cast<T>(sc.val[0]), dst, stream);
}
};
@@ -2292,14 +2292,12 @@ namespace
{
static void call(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream)
{
Scalar_<unsigned int> isc = sc;
unsigned int packedVal = 0;
packedVal |= (isc.val[0] & 0xffff);
packedVal |= (isc.val[1] & 0xffff) << 8;
packedVal |= (isc.val[2] & 0xffff) << 16;
packedVal |= (isc.val[3] & 0xffff) << 24;
packedVal |= (saturate_cast<unsigned char>(sc.val[0]) & 0xffff);
packedVal |= (saturate_cast<unsigned char>(sc.val[1]) & 0xffff) << 8;
packedVal |= (saturate_cast<unsigned char>(sc.val[2]) & 0xffff) << 16;
packedVal |= (saturate_cast<unsigned char>(sc.val[3]) & 0xffff) << 24;
func(src, packedVal, dst, stream);
}
@@ -2330,7 +2328,7 @@ namespace
oSizeROI.width = src.cols;
oSizeROI.height = src.rows;
const npp_t pConstants[] = {static_cast<npp_t>(sc.val[0]), static_cast<npp_t>(sc.val[1]), static_cast<npp_t>(sc.val[2]), static_cast<npp_t>(sc.val[3])};
const npp_t pConstants[] = {saturate_cast<npp_t>(sc.val[0]), saturate_cast<npp_t>(sc.val[1]), saturate_cast<npp_t>(sc.val[2]), saturate_cast<npp_t>(sc.val[3])};
nppSafeCall( func(src.ptr<npp_t>(), static_cast<int>(src.step), pConstants, dst.ptr<npp_t>(), static_cast<int>(dst.step), oSizeROI) );
@@ -2350,7 +2348,7 @@ namespace
oSizeROI.width = src.cols;
oSizeROI.height = src.rows;
nppSafeCall( func(src.ptr<npp_t>(), static_cast<int>(src.step), static_cast<npp_t>(sc.val[0]), dst.ptr<npp_t>(), static_cast<int>(dst.step), oSizeROI) );
nppSafeCall( func(src.ptr<npp_t>(), static_cast<int>(src.step), saturate_cast<npp_t>(sc.val[0]), dst.ptr<npp_t>(), static_cast<int>(dst.step), oSizeROI) );
if (stream == 0)
cudaSafeCall( cudaDeviceSynchronize() );
@@ -2365,11 +2363,11 @@ void cv::gpu::bitwise_and(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stre
typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[5][4] =
{
{BitScalar< bitScalarAnd<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiAndC_8u_C3R >::call, BitScalar4< bitScalarAnd<unsigned int> >::call},
{BitScalar<unsigned char, bitScalarAnd<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiAndC_8u_C3R >::call, BitScalar4< bitScalarAnd<unsigned int> >::call},
{0,0,0,0},
{BitScalar< bitScalarAnd<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiAndC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiAndC_16u_C4R>::call},
{BitScalar<unsigned short, bitScalarAnd<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiAndC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiAndC_16u_C4R>::call},
{0,0,0,0},
{BitScalar< bitScalarAnd<unsigned int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiAndC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiAndC_32s_C4R>::call}
{BitScalar<int, bitScalarAnd<int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiAndC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiAndC_32s_C4R>::call}
};
const int depth = src.depth();
@@ -2390,11 +2388,11 @@ void cv::gpu::bitwise_or(const GpuMat& src, const Scalar& sc, GpuMat& dst, Strea
typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[5][4] =
{
{BitScalar< bitScalarOr<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiOrC_8u_C3R >::call, BitScalar4< bitScalarOr<unsigned int> >::call},
{BitScalar<unsigned char, bitScalarOr<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiOrC_8u_C3R >::call, BitScalar4< bitScalarOr<unsigned int> >::call},
{0,0,0,0},
{BitScalar< bitScalarOr<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiOrC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiOrC_16u_C4R>::call},
{BitScalar<unsigned short, bitScalarOr<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiOrC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiOrC_16u_C4R>::call},
{0,0,0,0},
{BitScalar< bitScalarOr<unsigned int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiOrC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiOrC_32s_C4R>::call}
{BitScalar<int, bitScalarOr<int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiOrC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiOrC_32s_C4R>::call}
};
const int depth = src.depth();
@@ -2415,11 +2413,11 @@ void cv::gpu::bitwise_xor(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stre
typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[5][4] =
{
{BitScalar< bitScalarXor<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiXorC_8u_C3R >::call, BitScalar4< bitScalarXor<unsigned int> >::call},
{BitScalar<unsigned char, bitScalarXor<unsigned char> >::call , 0, NppBitwiseC<CV_8U , 3, nppiXorC_8u_C3R >::call, BitScalar4< bitScalarXor<unsigned int> >::call},
{0,0,0,0},
{BitScalar< bitScalarXor<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiXorC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiXorC_16u_C4R>::call},
{BitScalar<unsigned short, bitScalarXor<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiXorC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiXorC_16u_C4R>::call},
{0,0,0,0},
{BitScalar< bitScalarXor<unsigned int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiXorC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiXorC_32s_C4R>::call}
{BitScalar<int, bitScalarXor<int> >::call , 0, NppBitwiseC<CV_32S, 3, nppiXorC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiXorC_32s_C4R>::call}
};
const int depth = src.depth();

View File

@@ -104,12 +104,12 @@ void cv::gpu::connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scal
void cv::gpu::labelComponents(const GpuMat& mask, GpuMat& components, int flags, Stream& s)
{
if (!TargetArchs::builtWith(SHARED_ATOMICS) || !DeviceInfo().supports(SHARED_ATOMICS))
CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
CV_Assert(!mask.empty() && mask.type() == CV_8U);
if (mask.size() != components.size() || components.type() != CV_32SC1)
components.create(mask.size(), CV_32SC1);
if (!deviceSupports(SHARED_ATOMICS))
CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
components.create(mask.size(), CV_32SC1);
cudaStream_t stream = StreamAccessor::getStream(s);
device::ccl::labelComponents(mask, components, flags, stream);

View File

@@ -517,6 +517,7 @@ void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, d
CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
dst.create(dsize, src.type());
dst.setTo(Scalar::all(0));
funcs[src.depth()][src.channels() - 1](src, dst, dsize, angle, xShift, yShift, interpolation, StreamAccessor::getStream(stream));
}

View File

@@ -380,6 +380,7 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
dstcol[0] = static_cast<uchar>(sumcol[0] / comps.size[parent]);
dstcol[1] = static_cast<uchar>(sumcol[1] / comps.size[parent]);
dstcol[2] = static_cast<uchar>(sumcol[2] / comps.size[parent]);
dstcol[3] = 255;
}
}
}

View File

@@ -206,6 +206,8 @@ void cv::gpu::PyrLKOpticalFlow::dense(const GpuMat& prevImg, const GpuMat& nextI
ensureSizeIsEnough(prevImg.size(), CV_32FC1, vPyr_[0]);
ensureSizeIsEnough(prevImg.size(), CV_32FC1, uPyr_[1]);
ensureSizeIsEnough(prevImg.size(), CV_32FC1, vPyr_[1]);
uPyr_[0].setTo(Scalar::all(0));
vPyr_[0].setTo(Scalar::all(0));
uPyr_[1].setTo(Scalar::all(0));
vPyr_[1].setTo(Scalar::all(0));

View File

@@ -232,10 +232,8 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
};
bool useNpp = borderMode == BORDER_CONSTANT && ofs.x == 0 && ofs.y == 0 && useNppTab[src.depth()][src.channels() - 1][interpolation];
#ifdef linux
// NPP bug on float data
useNpp = useNpp && src.depth() != CV_32F;
#endif
// NPP bug on float data
useNpp = useNpp && src.depth() != CV_32F;
if (useNpp)
{
@@ -372,10 +370,8 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
};
bool useNpp = borderMode == BORDER_CONSTANT && ofs.x == 0 && ofs.y == 0 && useNppTab[src.depth()][src.channels() - 1][interpolation];
#ifdef linux
// NPP bug on float data
useNpp = useNpp && src.depth() != CV_32F;
#endif
// NPP bug on float data
useNpp = useNpp && src.depth() != CV_32F;
if (useNpp)
{