Skip to content

O
opencv
Commit 6397fa5b authored by Vladislav Vinogradov's avatar Vladislav Vinogradov
Browse files
Options

fixed gpu::pyrUp (now it matches cpu analog) 

fixed several warnings
parent 484fe1d5
Show whitespace changes
Inline Side-by-side
Showing with 466 additions and 415 deletions
modules/gpu/doc/image_processing.rst
View file @ 6397fa5b
...@@ -713,14 +713,12 @@ gpu::pyrDown ...@@ -713,14 +713,12 @@ gpu::pyrDown
------------------- -------------------
Smoothes an image and downsamples it. Smoothes an image and downsamples it.
.. ocv:function:: void gpu::pyrDown(const GpuMat& src, GpuMat& dst, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null()) .. ocv:function:: void gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
:param src: Source image. :param src: Source image.
:param dst: Destination image. Will have ``Size((src.cols+1)/2, (src.rows+1)/2)`` size and the same type as ``src`` . :param dst: Destination image. Will have ``Size((src.cols+1)/2, (src.rows+1)/2)`` size and the same type as ``src`` .
:param borderType: Pixel extrapolation method (see :ocv:func:`borderInterpolate` ). ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
:param stream: Stream for the asynchronous version. :param stream: Stream for the asynchronous version.
.. seealso:: :ocv:func:`pyrDown` .. seealso:: :ocv:func:`pyrDown`
...@@ -731,14 +729,12 @@ gpu::pyrUp ...@@ -731,14 +729,12 @@ gpu::pyrUp
------------------- -------------------
Upsamples an image and then smoothes it. Upsamples an image and then smoothes it.
.. ocv:function:: void gpu::pyrUp(const GpuMat& src, GpuMat& dst, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null()) .. ocv:function:: void gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
:param src: Source image. :param src: Source image.
:param dst: Destination image. Will have ``Size(src.cols*2, src.rows*2)`` size and the same type as ``src`` . :param dst: Destination image. Will have ``Size(src.cols*2, src.rows*2)`` size and the same type as ``src`` .
:param borderType: Pixel extrapolation method (see :ocv:func:`borderInterpolate` ). ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
:param stream: Stream for the asynchronous version. :param stream: Stream for the asynchronous version.
.. seealso:: :ocv:func:`pyrUp` .. seealso:: :ocv:func:`pyrUp`
......
modules/gpu/include/opencv2/gpu/gpu.hpp
View file @ 6397fa5b
...@@ -836,10 +836,10 @@ private: ...@@ -836,10 +836,10 @@ private:
CV_EXPORTS void matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, Stream& stream = Stream::Null()); CV_EXPORTS void matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, Stream& stream = Stream::Null());
//! smoothes the source image and downsamples it //! smoothes the source image and downsamples it
CV_EXPORTS void pyrDown(const GpuMat& src, GpuMat& dst, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null()); CV_EXPORTS void pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
//! upsamples the source image and then smoothes it //! upsamples the source image and then smoothes it
CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null()); CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
//! performs linear blending of two images //! performs linear blending of two images
//! to avoid accuracy errors sum of weigths shouldn't be very close to zero //! to avoid accuracy errors sum of weigths shouldn't be very close to zero
......
modules/gpu/src/cuda/pyr_down.cu
View file @ 6397fa5b
...@@ -124,7 +124,7 @@ namespace cv { namespace gpu { namespace device ...@@ -124,7 +124,7 @@ namespace cv { namespace gpu { namespace device
} }
} }
template <typename T, template <typename> class B> void pyrDown_caller(const DevMem2D_<T>& src, const DevMem2D_<T>& dst, cudaStream_t stream) template <typename T, template <typename> class B> void pyrDown_caller(DevMem2D_<T> src, DevMem2D_<T> dst, cudaStream_t stream)
{ {
const dim3 block(256); const dim3 block(256);
const dim3 grid(divUp(src.cols, block.x), dst.rows); const dim3 grid(divUp(src.cols, block.x), dst.rows);
...@@ -138,48 +138,39 @@ namespace cv { namespace gpu { namespace device ...@@ -138,48 +138,39 @@ namespace cv { namespace gpu { namespace device
cudaSafeCall( cudaDeviceSynchronize() ); cudaSafeCall( cudaDeviceSynchronize() );
} }
template <typename T, int cn> void pyrDown_gpu(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream) template <typename T> void pyrDown_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream)
{ {
typedef typename TypeVec<T, cn>::vec_type type; pyrDown_caller<T, BrdReflect101>(static_cast< DevMem2D_<T> >(src), static_cast< DevMem2D_<T> >(dst), stream);
typedef void (*caller_t)(const DevMem2D_<type>& src, const DevMem2D_<type>& dst, cudaStream_t stream);
static const caller_t callers[] =
{
pyrDown_caller<type, BrdReflect101>, pyrDown_caller<type, BrdReplicate>, pyrDown_caller<type, BrdConstant>, pyrDown_caller<type, BrdReflect>, pyrDown_caller<type, BrdWrap>
};
callers[borderType](static_cast< DevMem2D_<type> >(src), static_cast< DevMem2D_<type> >(dst), stream);
} }
template void pyrDown_gpu<uchar, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<uchar>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<uchar, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<uchar2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<uchar, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<uchar3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<uchar, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<uchar4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<schar, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<schar>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<schar, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<char2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<schar, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<char3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<schar, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<char4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<ushort, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<ushort>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<ushort, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<ushort2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<ushort, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<ushort3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<ushort, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<ushort4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<short, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<short>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<short, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<short2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<short, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<short3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<short, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<short4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<int, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<int>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<int, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<int2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<int, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<int3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<int, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<int4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<float, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<float>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<float, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrDown_gpu<float2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<float, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<float3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrDown_gpu<float, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrDown_gpu<float4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
} // namespace imgproc } // namespace imgproc
}}} // namespace cv { namespace gpu { namespace device }}} // namespace cv { namespace gpu { namespace device
modules/gpu/src/cuda/pyr_up.cu
View file @ 6397fa5b
...@@ -50,10 +50,9 @@ namespace cv { namespace gpu { namespace device ...@@ -50,10 +50,9 @@ namespace cv { namespace gpu { namespace device
{ {
namespace imgproc namespace imgproc
{ {
template <class SrcPtr, typename D> __global__ void pyrUp(const SrcPtr src, DevMem2D_<D> dst) template <typename T> __global__ void pyrUp(const DevMem2D_<T> src, DevMem2D_<T> dst)
{ {
typedef typename SrcPtr::elem_type src_t; typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type sum_t;
typedef typename TypeVec<float, VecTraits<D>::cn>::vec_type sum_t;
const int x = blockIdx.x * blockDim.x + threadIdx.x; const int x = blockIdx.x * blockDim.x + threadIdx.x;
const int y = blockIdx.y * blockDim.y + threadIdx.y; const int y = blockIdx.y * blockDim.y + threadIdx.y;
...@@ -63,8 +62,14 @@ namespace cv { namespace gpu { namespace device ...@@ -63,8 +62,14 @@ namespace cv { namespace gpu { namespace device
if (threadIdx.x < 10 && threadIdx.y < 10) if (threadIdx.x < 10 && threadIdx.y < 10)
{ {
const int srcx = static_cast<int>((blockIdx.x * blockDim.x) / 2 + threadIdx.x) - 1; int srcx = static_cast<int>((blockIdx.x * blockDim.x) / 2 + threadIdx.x) - 1;
const int srcy = static_cast<int>((blockIdx.y * blockDim.y) / 2 + threadIdx.y) - 1; int srcy = static_cast<int>((blockIdx.y * blockDim.y) / 2 + threadIdx.y) - 1;
srcx = ::abs(srcx);
srcx = ::min(src.cols - 1, srcx);
srcy = ::abs(srcy);
srcy = ::min(src.rows - 1, srcy);
s_srcPatch[threadIdx.y][threadIdx.x] = saturate_cast<sum_t>(src(srcy, srcx)); s_srcPatch[threadIdx.y][threadIdx.x] = saturate_cast<sum_t>(src(srcy, srcx));
} }
...@@ -134,66 +139,54 @@ namespace cv { namespace gpu { namespace device ...@@ -134,66 +139,54 @@ namespace cv { namespace gpu { namespace device
sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][threadIdx.x]; sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][threadIdx.x];
if (x < dst.cols && y < dst.rows) if (x < dst.cols && y < dst.rows)
dst(y, x) = saturate_cast<D>(4.0f * sum); dst(y, x) = saturate_cast<T>(4.0f * sum);
} }
template <typename T, template <typename> class B> void pyrUp_caller(const DevMem2D_<T>& src, const DevMem2D_<T>& dst, cudaStream_t stream) template <typename T> void pyrUp_caller(DevMem2D_<T> src, DevMem2D_<T> dst, cudaStream_t stream)
{ {
const dim3 block(16, 16); const dim3 block(16, 16);
const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
B<T> b(src.rows, src.cols); pyrUp<<<grid, block, 0, stream>>>(src, dst);
BorderReader< PtrStep<T>, B<T> > srcReader(src, b);
pyrUp<<<grid, block, 0, stream>>>(srcReader, dst);
cudaSafeCall( cudaGetLastError() ); cudaSafeCall( cudaGetLastError() );
if (stream == 0) if (stream == 0)
cudaSafeCall( cudaDeviceSynchronize() ); cudaSafeCall( cudaDeviceSynchronize() );
} }
template <typename T, int cn> void pyrUp_gpu(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream) template <typename T> void pyrUp_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream)
{ {
typedef typename TypeVec<T, cn>::vec_type type; pyrUp_caller<T>(static_cast< DevMem2D_<T> >(src), static_cast< DevMem2D_<T> >(dst), stream);
typedef void (*caller_t)(const DevMem2D_<type>& src, const DevMem2D_<type>& dst, cudaStream_t stream);
static const caller_t callers[] =
{
pyrUp_caller<type, BrdReflect101>, pyrUp_caller<type, BrdReplicate>, pyrUp_caller<type, BrdConstant>, pyrUp_caller<type, BrdReflect>, pyrUp_caller<type, BrdWrap>
};
callers[borderType](static_cast< DevMem2D_<type> >(src), static_cast< DevMem2D_<type> >(dst), stream);
} }
template void pyrUp_gpu<uchar, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<uchar>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<uchar, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<uchar2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<uchar, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<uchar3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<uchar, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<uchar4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<schar, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<schar>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<schar, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<char2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<schar, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<char3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<schar, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<char4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<ushort, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<ushort>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<ushort, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<ushort2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<ushort, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<ushort3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<ushort, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<ushort4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<short, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<short>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<short, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<short2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<short, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<short3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<short, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<short4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<int, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<int>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<int, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<int2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<int, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<int3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<int, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<int4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<float, 1>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<float>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<float, 2>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); //template void pyrUp_gpu<float2>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<float, 3>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<float3>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template void pyrUp_gpu<float, 4>(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream); template void pyrUp_gpu<float4>(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
} // namespace imgproc } // namespace imgproc
}}} // namespace cv { namespace gpu { namespace device }}} // namespace cv { namespace gpu { namespace device
modules/gpu/src/imgproc.cpp
View file @ 6397fa5b
...@@ -87,8 +87,6 @@ void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int, Stream&) { throw_nogpu(); } ...@@ -87,8 +87,6 @@ void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int, Stream&) { throw_nogpu(); }
void cv::gpu::ConvolveBuf::create(Size, Size) { throw_nogpu(); } void cv::gpu::ConvolveBuf::create(Size, Size) { throw_nogpu(); }
void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool) { throw_nogpu(); } void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool) { throw_nogpu(); }
void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&, Stream&) { throw_nogpu(); } void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&, Stream&) { throw_nogpu(); }
void cv::gpu::pyrDown(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
void cv::gpu::pyrUp(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int, bool) { throw_nogpu(); } void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
void cv::gpu::Canny(const GpuMat&, CannyBuf&, GpuMat&, double, double, int, bool) { throw_nogpu(); } void cv::gpu::Canny(const GpuMat&, CannyBuf&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
void cv::gpu::Canny(const GpuMat&, const GpuMat&, GpuMat&, double, double, bool) { throw_nogpu(); } void cv::gpu::Canny(const GpuMat&, const GpuMat&, GpuMat&, double, double, bool) { throw_nogpu(); }
...@@ -96,8 +94,6 @@ void cv::gpu::Canny(const GpuMat&, const GpuMat&, CannyBuf&, GpuMat&, double, do ...@@ -96,8 +94,6 @@ void cv::gpu::Canny(const GpuMat&, const GpuMat&, CannyBuf&, GpuMat&, double, do
cv::gpu::CannyBuf::CannyBuf(const GpuMat&, const GpuMat&) { throw_nogpu(); } cv::gpu::CannyBuf::CannyBuf(const GpuMat&, const GpuMat&) { throw_nogpu(); }
void cv::gpu::CannyBuf::create(const Size&, int) { throw_nogpu(); } void cv::gpu::CannyBuf::create(const Size&, int) { throw_nogpu(); }
void cv::gpu::CannyBuf::release() { throw_nogpu(); } void cv::gpu::CannyBuf::release() { throw_nogpu(); }
void cv::gpu::ImagePyramid::build(const GpuMat&, int, Stream&) { throw_nogpu(); }
void cv::gpu::ImagePyramid::getLayer(GpuMat&, Size, Stream&) const { throw_nogpu(); }
#else /* !defined (HAVE_CUDA) */ #else /* !defined (HAVE_CUDA) */
...@@ -1421,83 +1417,6 @@ void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, ...@@ -1421,83 +1417,6 @@ void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
#endif #endif
} }
//////////////////////////////////////////////////////////////////////////////
// pyrDown
namespace cv { namespace gpu { namespace device
{
namespace imgproc
{
template <typename T, int cn> void pyrDown_gpu(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream);
}
}}}
void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, int borderType, Stream& stream)
{
using namespace ::cv::gpu::device::imgproc;
typedef void (*func_t)(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{pyrDown_gpu<uchar, 1>, pyrDown_gpu<uchar, 2>, pyrDown_gpu<uchar, 3>, pyrDown_gpu<uchar, 4>},
{pyrDown_gpu<schar, 1>, pyrDown_gpu<schar, 2>, pyrDown_gpu<schar, 3>, pyrDown_gpu<schar, 4>},
{pyrDown_gpu<ushort, 1>, pyrDown_gpu<ushort, 2>, pyrDown_gpu<ushort, 3>, pyrDown_gpu<ushort, 4>},
{pyrDown_gpu<short, 1>, pyrDown_gpu<short, 2>, pyrDown_gpu<short, 3>, pyrDown_gpu<short, 4>},
{pyrDown_gpu<int, 1>, pyrDown_gpu<int, 2>, pyrDown_gpu<int, 3>, pyrDown_gpu<int, 4>},
{pyrDown_gpu<float, 1>, pyrDown_gpu<float, 2>, pyrDown_gpu<float, 3>, pyrDown_gpu<float, 4>},
};
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
CV_Assert(borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP);
int gpuBorderType;
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
funcs[src.depth()][src.channels() - 1](src, dst, gpuBorderType, StreamAccessor::getStream(stream));
}
//////////////////////////////////////////////////////////////////////////////
// pyrUp
namespace cv { namespace gpu { namespace device
{
namespace imgproc
{
template <typename T, int cn> void pyrUp_gpu(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream);
}
}}}
void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, int borderType, Stream& stream)
{
using namespace ::cv::gpu::device::imgproc;
typedef void (*func_t)(const DevMem2Db& src, const DevMem2Db& dst, int borderType, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{pyrUp_gpu<uchar, 1>, pyrUp_gpu<uchar, 2>, pyrUp_gpu<uchar, 3>, pyrUp_gpu<uchar, 4>},
{pyrUp_gpu<schar, 1>, pyrUp_gpu<schar, 2>, pyrUp_gpu<schar, 3>, pyrUp_gpu<schar, 4>},
{pyrUp_gpu<ushort, 1>, pyrUp_gpu<ushort, 2>, pyrUp_gpu<ushort, 3>, pyrUp_gpu<ushort, 4>},
{pyrUp_gpu<short, 1>, pyrUp_gpu<short, 2>, pyrUp_gpu<short, 3>, pyrUp_gpu<short, 4>},
{pyrUp_gpu<int, 1>, pyrUp_gpu<int, 2>, pyrUp_gpu<int, 3>, pyrUp_gpu<int, 4>},
{pyrUp_gpu<float, 1>, pyrUp_gpu<float, 2>, pyrUp_gpu<float, 3>, pyrUp_gpu<float, 4>},
};
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
CV_Assert(borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP);
int gpuBorderType;
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
dst.create(src.rows*2, src.cols*2, src.type());
funcs[src.depth()][src.channels() - 1](src, dst, gpuBorderType, StreamAccessor::getStream(stream));
}
////////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////////
// Canny // Canny
...@@ -1646,129 +1565,6 @@ void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& d ...@@ -1646,129 +1565,6 @@ void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& d
CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh)); CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
} }
//////////////////////////////////////////////////////////////////////////////
// ImagePyramid
namespace cv { namespace gpu { namespace device
{
namespace pyramid
{
template <typename T> void kernelDownsampleX2_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template <typename T> void kernelInterpolateFrom1_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
}
}}}
void cv::gpu::ImagePyramid::build(const GpuMat& img, int numLayers, Stream& stream)
{
using namespace cv::gpu::device::pyramid;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[7][4] =
{
{kernelDownsampleX2_gpu<uchar1>, /*kernelDownsampleX2_gpu<uchar2>*/ 0, kernelDownsampleX2_gpu<uchar3>, kernelDownsampleX2_gpu<uchar4>},
{/*kernelDownsampleX2_gpu<char1>*/0, /*kernelDownsampleX2_gpu<char2>*/ 0, /*kernelDownsampleX2_gpu<char3>*/ 0, /*kernelDownsampleX2_gpu<char4>*/ 0},
{kernelDownsampleX2_gpu<ushort1>, /*kernelDownsampleX2_gpu<ushort2>*/ 0, kernelDownsampleX2_gpu<ushort3>, kernelDownsampleX2_gpu<ushort4>},
{/*kernelDownsampleX2_gpu<short1>*/ 0, /*kernelDownsampleX2_gpu<short2>*/ 0, /*kernelDownsampleX2_gpu<short3>*/ 0, /*kernelDownsampleX2_gpu<short4>*/ 0},
{/*kernelDownsampleX2_gpu<int1>*/ 0, /*kernelDownsampleX2_gpu<int2>*/ 0, /*kernelDownsampleX2_gpu<int3>*/ 0, /*kernelDownsampleX2_gpu<int4>*/ 0},
{kernelDownsampleX2_gpu<float1>, /*kernelDownsampleX2_gpu<float2>*/ 0, kernelDownsampleX2_gpu<float3>, kernelDownsampleX2_gpu<float4>},
{/*kernelDownsampleX2_gpu<double1>*/ 0, /*kernelDownsampleX2_gpu<double2>*/ 0, /*kernelDownsampleX2_gpu<double3>*/ 0, /*kernelDownsampleX2_gpu<double4>*/ 0}
};
CV_Assert(img.channels() == 1 || img.channels() == 3 || img.channels() == 4);
CV_Assert(img.depth() == CV_8U || img.depth() == CV_16U || img.depth() == CV_32F);
layer0_ = img;
Size szLastLayer = img.size();
nLayers_ = 1;
if (numLayers <= 0)
numLayers = 255; //it will cut-off when any of the dimensions goes 1
pyramid_.resize(numLayers);
for (int i = 0; i < numLayers - 1; ++i)
{
Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
if (szCurLayer.width == 0 || szCurLayer.height == 0)
break;
ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
nLayers_++;
const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
func_t func = funcs[img.depth()][img.channels() - 1];
CV_Assert(func != 0);
func(prevLayer, pyramid_[i], StreamAccessor::getStream(stream));
szLastLayer = szCurLayer;
}
}
void cv::gpu::ImagePyramid::getLayer(GpuMat& outImg, Size outRoi, Stream& stream) const
{
using namespace cv::gpu::device::pyramid;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[7][4] =
{
{kernelInterpolateFrom1_gpu<uchar1>, /*kernelInterpolateFrom1_gpu<uchar2>*/ 0, kernelInterpolateFrom1_gpu<uchar3>, kernelInterpolateFrom1_gpu<uchar4>},
{/*kernelInterpolateFrom1_gpu<char1>*/0, /*kernelInterpolateFrom1_gpu<char2>*/ 0, /*kernelInterpolateFrom1_gpu<char3>*/ 0, /*kernelInterpolateFrom1_gpu<char4>*/ 0},
{kernelInterpolateFrom1_gpu<ushort1>, /*kernelInterpolateFrom1_gpu<ushort2>*/ 0, kernelInterpolateFrom1_gpu<ushort3>, kernelInterpolateFrom1_gpu<ushort4>},
{/*kernelInterpolateFrom1_gpu<short1>*/ 0, /*kernelInterpolateFrom1_gpu<short2>*/ 0, /*kernelInterpolateFrom1_gpu<short3>*/ 0, /*kernelInterpolateFrom1_gpu<short4>*/ 0},
{/*kernelInterpolateFrom1_gpu<int1>*/ 0, /*kernelInterpolateFrom1_gpu<int2>*/ 0, /*kernelInterpolateFrom1_gpu<int3>*/ 0, /*kernelInterpolateFrom1_gpu<int4>*/ 0},
{kernelInterpolateFrom1_gpu<float1>, /*kernelInterpolateFrom1_gpu<float2>*/ 0, kernelInterpolateFrom1_gpu<float3>, kernelInterpolateFrom1_gpu<float4>},
{/*kernelInterpolateFrom1_gpu<double1>*/ 0, /*kernelInterpolateFrom1_gpu<double2>*/ 0, /*kernelInterpolateFrom1_gpu<double3>*/ 0, /*kernelInterpolateFrom1_gpu<double4>*/ 0}
};
CV_Assert(outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0);
ensureSizeIsEnough(outRoi, layer0_.type(), outImg);
if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
{
if (stream)
stream.enqueueCopy(layer0_, outImg);
else
layer0_.copyTo(outImg);
}
float lastScale = 1.0f;
float curScale;
GpuMat lastLayer = layer0_;
GpuMat curLayer;
for (int i = 0; i < nLayers_ - 1; ++i)
{
curScale = lastScale * 0.5f;
curLayer = pyramid_[i];
if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
{
if (stream)
stream.enqueueCopy(curLayer, outImg);
else
curLayer.copyTo(outImg);
}
if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
break;
lastScale = curScale;
lastLayer = curLayer;
}
func_t func = funcs[outImg.depth()][outImg.channels() - 1];
CV_Assert(func != 0);
func(lastLayer, outImg, StreamAccessor::getStream(stream));
}
#endif /* !defined (HAVE_CUDA) */ #endif /* !defined (HAVE_CUDA) */
modules/gpu/src/pyramids.cpp 0 → 100644
View file @ 6397fa5b
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
#ifndef HAVE_CUDA
void cv::gpu::pyrDown(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
void cv::gpu::pyrUp(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
void cv::gpu::ImagePyramid::build(const GpuMat&, int, Stream&) { throw_nogpu(); }
void cv::gpu::ImagePyramid::getLayer(GpuMat&, Size, Stream&) const { throw_nogpu(); }
#else // HAVE_CUDA
//////////////////////////////////////////////////////////////////////////////
// pyrDown
namespace cv { namespace gpu { namespace device
{
namespace imgproc
{
template <typename T> void pyrDown_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
}
}}}
void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream)
{
using namespace cv::gpu::device::imgproc;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{pyrDown_gpu<uchar> , 0 /*pyrDown_gpu<uchar2>*/ , pyrDown_gpu<uchar3> , pyrDown_gpu<uchar4> },
{0 /*pyrDown_gpu<schar>*/, 0 /*pyrDown_gpu<schar2>*/ , 0 /*pyrDown_gpu<schar3>*/, 0 /*pyrDown_gpu<schar4>*/},
{pyrDown_gpu<ushort> , 0 /*pyrDown_gpu<ushort2>*/, pyrDown_gpu<ushort3> , pyrDown_gpu<ushort4> },
{pyrDown_gpu<short> , 0 /*pyrDown_gpu<short2>*/ , pyrDown_gpu<short3> , pyrDown_gpu<short4> },
{0 /*pyrDown_gpu<int>*/ , 0 /*pyrDown_gpu<int2>*/ , 0 /*pyrDown_gpu<int3>*/ , 0 /*pyrDown_gpu<int4>*/ },
{pyrDown_gpu<float> , 0 /*pyrDown_gpu<float2>*/ , pyrDown_gpu<float3> , pyrDown_gpu<float4> }
};
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
const func_t func = funcs[src.depth()][src.channels() - 1];
CV_Assert(func != 0);
dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
func(src, dst, StreamAccessor::getStream(stream));
}
//////////////////////////////////////////////////////////////////////////////
// pyrUp
namespace cv { namespace gpu { namespace device
{
namespace imgproc
{
template <typename T> void pyrUp_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
}
}}}
void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream)
{
using namespace cv::gpu::device::imgproc;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{pyrUp_gpu<uchar> , 0 /*pyrUp_gpu<uchar2>*/ , pyrUp_gpu<uchar3> , pyrUp_gpu<uchar4> },
{0 /*pyrUp_gpu<schar>*/, 0 /*pyrUp_gpu<schar2>*/ , 0 /*pyrUp_gpu<schar3>*/, 0 /*pyrUp_gpu<schar4>*/},
{pyrUp_gpu<ushort> , 0 /*pyrUp_gpu<ushort2>*/, pyrUp_gpu<ushort3> , pyrUp_gpu<ushort4> },
{pyrUp_gpu<short> , 0 /*pyrUp_gpu<short2>*/ , pyrUp_gpu<short3> , pyrUp_gpu<short4> },
{0 /*pyrUp_gpu<int>*/ , 0 /*pyrUp_gpu<int2>*/ , 0 /*pyrUp_gpu<int3>*/ , 0 /*pyrUp_gpu<int4>*/ },
{pyrUp_gpu<float> , 0 /*pyrUp_gpu<float2>*/ , pyrUp_gpu<float3> , pyrUp_gpu<float4> }
};
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
const func_t func = funcs[src.depth()][src.channels() - 1];
CV_Assert(func != 0);
dst.create(src.rows * 2, src.cols * 2, src.type());
func(src, dst, StreamAccessor::getStream(stream));
}
//////////////////////////////////////////////////////////////////////////////
// ImagePyramid
namespace cv { namespace gpu { namespace device
{
namespace pyramid
{
template <typename T> void kernelDownsampleX2_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
template <typename T> void kernelInterpolateFrom1_gpu(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
}
}}}
void cv::gpu::ImagePyramid::build(const GpuMat& img, int numLayers, Stream& stream)
{
using namespace cv::gpu::device::pyramid;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{kernelDownsampleX2_gpu<uchar1> , 0 /*kernelDownsampleX2_gpu<uchar2>*/ , kernelDownsampleX2_gpu<uchar3> , kernelDownsampleX2_gpu<uchar4> },
{0 /*kernelDownsampleX2_gpu<char1>*/ , 0 /*kernelDownsampleX2_gpu<char2>*/ , 0 /*kernelDownsampleX2_gpu<char3>*/ , 0 /*kernelDownsampleX2_gpu<char4>*/ },
{kernelDownsampleX2_gpu<ushort1> , 0 /*kernelDownsampleX2_gpu<ushort2>*/, kernelDownsampleX2_gpu<ushort3> , kernelDownsampleX2_gpu<ushort4> },
{0 /*kernelDownsampleX2_gpu<short1>*/ , 0 /*kernelDownsampleX2_gpu<short2>*/ , 0 /*kernelDownsampleX2_gpu<short3>*/, 0 /*kernelDownsampleX2_gpu<short4>*/},
{0 /*kernelDownsampleX2_gpu<int1>*/ , 0 /*kernelDownsampleX2_gpu<int2>*/ , 0 /*kernelDownsampleX2_gpu<int3>*/ , 0 /*kernelDownsampleX2_gpu<int4>*/ },
{kernelDownsampleX2_gpu<float1> , 0 /*kernelDownsampleX2_gpu<float2>*/ , kernelDownsampleX2_gpu<float3> , kernelDownsampleX2_gpu<float4> }
};
CV_Assert(img.depth() <= CV_32F && img.channels() <= 4);
const func_t func = funcs[img.depth()][img.channels() - 1];
CV_Assert(func != 0);
layer0_ = img;
Size szLastLayer = img.size();
nLayers_ = 1;
if (numLayers <= 0)
numLayers = 255; //it will cut-off when any of the dimensions goes 1
pyramid_.resize(numLayers);
for (int i = 0; i < numLayers - 1; ++i)
{
Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
if (szCurLayer.width == 0 || szCurLayer.height == 0)
break;
ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
nLayers_++;
const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
func(prevLayer, pyramid_[i], StreamAccessor::getStream(stream));
szLastLayer = szCurLayer;
}
}
void cv::gpu::ImagePyramid::getLayer(GpuMat& outImg, Size outRoi, Stream& stream) const
{
using namespace cv::gpu::device::pyramid;
typedef void (*func_t)(DevMem2Db src, DevMem2Db dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{kernelInterpolateFrom1_gpu<uchar1> , 0 /*kernelInterpolateFrom1_gpu<uchar2>*/ , kernelInterpolateFrom1_gpu<uchar3> , kernelInterpolateFrom1_gpu<uchar4> },
{0 /*kernelInterpolateFrom1_gpu<char1>*/ , 0 /*kernelInterpolateFrom1_gpu<char2>*/ , 0 /*kernelInterpolateFrom1_gpu<char3>*/ , 0 /*kernelInterpolateFrom1_gpu<char4>*/ },
{kernelInterpolateFrom1_gpu<ushort1> , 0 /*kernelInterpolateFrom1_gpu<ushort2>*/, kernelInterpolateFrom1_gpu<ushort3> , kernelInterpolateFrom1_gpu<ushort4> },
{0 /*kernelInterpolateFrom1_gpu<short1>*/, 0 /*kernelInterpolateFrom1_gpu<short2>*/ , 0 /*kernelInterpolateFrom1_gpu<short3>*/, 0 /*kernelInterpolateFrom1_gpu<short4>*/},
{0 /*kernelInterpolateFrom1_gpu<int1>*/ , 0 /*kernelInterpolateFrom1_gpu<int2>*/ , 0 /*kernelInterpolateFrom1_gpu<int3>*/ , 0 /*kernelInterpolateFrom1_gpu<int4>*/ },
{kernelInterpolateFrom1_gpu<float1> , 0 /*kernelInterpolateFrom1_gpu<float2>*/ , kernelInterpolateFrom1_gpu<float3> , kernelInterpolateFrom1_gpu<float4> }
};
CV_Assert(outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0);
ensureSizeIsEnough(outRoi, layer0_.type(), outImg);
const func_t func = funcs[outImg.depth()][outImg.channels() - 1];
CV_Assert(func != 0);
if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
{
if (stream)
stream.enqueueCopy(layer0_, outImg);
else
layer0_.copyTo(outImg);
}
float lastScale = 1.0f;
float curScale;
GpuMat lastLayer = layer0_;
GpuMat curLayer;
for (int i = 0; i < nLayers_ - 1; ++i)
{
curScale = lastScale * 0.5f;
curLayer = pyramid_[i];
if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
{
if (stream)
stream.enqueueCopy(curLayer, outImg);
else
curLayer.copyTo(outImg);
}
if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
break;
lastScale = curScale;
lastLayer = curLayer;
}
func(lastLayer, outImg, StreamAccessor::getStream(stream));
}
#endif // HAVE_CUDA
modules/gpu/test/test_calib3d.cpp
View file @ 6397fa5b
...@@ -230,7 +230,7 @@ TEST_P(ProjectPoints, Accuracy) ...@@ -230,7 +230,7 @@ TEST_P(ProjectPoints, Accuracy)
d_dst.download(dst); d_dst.download(dst);
ASSERT_EQ(dst_gold.size(), dst.cols); ASSERT_EQ(dst_gold.size(), static_cast<size_t>(dst.cols));
ASSERT_EQ(1, dst.rows); ASSERT_EQ(1, dst.rows);
ASSERT_EQ(CV_32FC2, dst.type()); ASSERT_EQ(CV_32FC2, dst.type());
......
modules/gpu/test/test_features2d.cpp
View file @ 6397fa5b
...@@ -218,7 +218,7 @@ TEST_P(BruteForceMatcher, Match) ...@@ -218,7 +218,7 @@ TEST_P(BruteForceMatcher, Match)
matcher.match(loadMat(query), loadMat(train), matches); matcher.match(loadMat(query), loadMat(train), matches);
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
for (size_t i = 0; i < matches.size(); i++) for (size_t i = 0; i < matches.size(); i++)
...@@ -259,7 +259,7 @@ TEST_P(BruteForceMatcher, MatchAdd) ...@@ -259,7 +259,7 @@ TEST_P(BruteForceMatcher, MatchAdd)
isMaskSupported = matcher.isMaskSupported(); isMaskSupported = matcher.isMaskSupported();
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
for (size_t i = 0; i < matches.size(); i++) for (size_t i = 0; i < matches.size(); i++)
...@@ -292,7 +292,7 @@ TEST_P(BruteForceMatcher, KnnMatch2) ...@@ -292,7 +292,7 @@ TEST_P(BruteForceMatcher, KnnMatch2)
cv::gpu::BruteForceMatcher_GPU_base matcher(distType); cv::gpu::BruteForceMatcher_GPU_base matcher(distType);
matcher.knnMatch(loadMat(query), loadMat(train), matches, knn); matcher.knnMatch(loadMat(query), loadMat(train), matches, knn);
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
for (size_t i = 0; i < matches.size(); i++) for (size_t i = 0; i < matches.size(); i++)
...@@ -324,7 +324,7 @@ TEST_P(BruteForceMatcher, KnnMatch3) ...@@ -324,7 +324,7 @@ TEST_P(BruteForceMatcher, KnnMatch3)
cv::gpu::BruteForceMatcher_GPU_base matcher(distType); cv::gpu::BruteForceMatcher_GPU_base matcher(distType);
matcher.knnMatch(loadMat(query), loadMat(train), matches, knn); matcher.knnMatch(loadMat(query), loadMat(train), matches, knn);
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
for (size_t i = 0; i < matches.size(); i++) for (size_t i = 0; i < matches.size(); i++)
...@@ -375,7 +375,7 @@ TEST_P(BruteForceMatcher, KnnMatchAdd2) ...@@ -375,7 +375,7 @@ TEST_P(BruteForceMatcher, KnnMatchAdd2)
isMaskSupported = matcher.isMaskSupported(); isMaskSupported = matcher.isMaskSupported();
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
int shift = isMaskSupported ? 1 : 0; int shift = isMaskSupported ? 1 : 0;
...@@ -437,7 +437,7 @@ TEST_P(BruteForceMatcher, KnnMatchAdd3) ...@@ -437,7 +437,7 @@ TEST_P(BruteForceMatcher, KnnMatchAdd3)
isMaskSupported = matcher.isMaskSupported(); isMaskSupported = matcher.isMaskSupported();
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
int shift = isMaskSupported ? 1 : 0; int shift = isMaskSupported ? 1 : 0;
...@@ -485,7 +485,7 @@ TEST_P(BruteForceMatcher, RadiusMatch) ...@@ -485,7 +485,7 @@ TEST_P(BruteForceMatcher, RadiusMatch)
matcher.radiusMatch(loadMat(query), loadMat(train), matches, radius); matcher.radiusMatch(loadMat(query), loadMat(train), matches, radius);
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
for (size_t i = 0; i < matches.size(); i++) for (size_t i = 0; i < matches.size(); i++)
...@@ -536,7 +536,7 @@ TEST_P(BruteForceMatcher, RadiusMatchAdd) ...@@ -536,7 +536,7 @@ TEST_P(BruteForceMatcher, RadiusMatchAdd)
isMaskSupported = matcher.isMaskSupported(); isMaskSupported = matcher.isMaskSupported();
ASSERT_EQ(queryDescCount, matches.size()); ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
int badCount = 0; int badCount = 0;
int shift = isMaskSupported ? 1 : 0; int shift = isMaskSupported ? 1 : 0;
...@@ -598,7 +598,6 @@ struct FAST : TestWithParam<cv::gpu::DeviceInfo> ...@@ -598,7 +598,6 @@ struct FAST : TestWithParam<cv::gpu::DeviceInfo>
image = readImage("features2d/aloe.png", CV_LOAD_IMAGE_GRAYSCALE); image = readImage("features2d/aloe.png", CV_LOAD_IMAGE_GRAYSCALE);
ASSERT_FALSE(image.empty()); ASSERT_FALSE(image.empty());
cv::RNG& rng = cvtest::TS::ptr()->get_rng();
threshold = 30; threshold = 30;
cv::FAST(image, keypoints_gold, threshold); cv::FAST(image, keypoints_gold, threshold);
......
modules/gpu/test/test_imgproc.cpp
View file @ 6397fa5b
...@@ -2279,8 +2279,6 @@ PARAM_TEST_CASE(CornerMinEigen, cv::gpu::DeviceInfo, MatType, Border, int, int) ...@@ -2279,8 +2279,6 @@ PARAM_TEST_CASE(CornerMinEigen, cv::gpu::DeviceInfo, MatType, Border, int, int)
cv::gpu::setDevice(devInfo.deviceID()); cv::gpu::setDevice(devInfo.deviceID());
cv::RNG& rng = TS::ptr()->get_rng();
cv::Mat img = readImage("stereobm/aloe-L.png", CV_LOAD_IMAGE_GRAYSCALE); cv::Mat img = readImage("stereobm/aloe-L.png", CV_LOAD_IMAGE_GRAYSCALE);
ASSERT_FALSE(img.empty()); ASSERT_FALSE(img.empty());
...@@ -3101,105 +3099,6 @@ INSTANTIATE_TEST_CASE_P(ImgProc, Blend, Combine( ...@@ -3101,105 +3099,6 @@ INSTANTIATE_TEST_CASE_P(ImgProc, Blend, Combine(
testing::Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4), testing::Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
WHOLE_SUBMAT)); WHOLE_SUBMAT));
////////////////////////////////////////////////////////
// pyrDown
PARAM_TEST_CASE(PyrDown, cv::gpu::DeviceInfo, MatType, UseRoi)
{
cv::gpu::DeviceInfo devInfo;
int type;
bool useRoi;
cv::Mat src;
cv::Mat dst_gold;
virtual void SetUp()
{
devInfo = GET_PARAM(0);
type = GET_PARAM(1);
useRoi = GET_PARAM(2);
cv::gpu::setDevice(devInfo.deviceID());
cv::RNG& rng = TS::ptr()->get_rng();
cv::Size size(rng.uniform(100, 200), rng.uniform(100, 200));
src = randomMat(rng, size, type, 0.0, 255.0, false);
cv::pyrDown(src, dst_gold);
}
};
TEST_P(PyrDown, Accuracy)
{
cv::Mat dst;
cv::gpu::GpuMat d_dst;
cv::gpu::pyrDown(loadMat(src, useRoi), d_dst);
d_dst.download(dst);
EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-4 : 1.0);
}
INSTANTIATE_TEST_CASE_P(ImgProc, PyrDown, Combine(
ALL_DEVICES,
Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_16UC1, CV_16UC3, CV_16UC4, CV_32FC1, CV_32FC3, CV_32FC4),
WHOLE_SUBMAT));
////////////////////////////////////////////////////////
// pyrUp
PARAM_TEST_CASE(PyrUp, cv::gpu::DeviceInfo, MatType, UseRoi)
{
cv::gpu::DeviceInfo devInfo;
int type;
bool useRoi;
cv::Mat src;
cv::Mat dst_gold;
virtual void SetUp()
{
devInfo = GET_PARAM(0);
type = GET_PARAM(1);
useRoi = GET_PARAM(2);
cv::gpu::setDevice(devInfo.deviceID());
cv::RNG& rng = TS::ptr()->get_rng();
cv::Size size(rng.uniform(200, 400), rng.uniform(200, 400));
src = randomMat(rng, size, type, 0.0, 255.0, false);
cv::pyrUp(src, dst_gold);
}
};
TEST_P(PyrUp, Accuracy)
{
cv::Mat dst;
cv::gpu::GpuMat d_dst;
cv::gpu::pyrUp(loadMat(src, useRoi), d_dst, cv::BORDER_REFLECT);
d_dst.download(dst);
// results differs only on border left and top border due different border extrapolation type
EXPECT_MAT_NEAR(dst_gold(cv::Range(1, dst_gold.rows), cv::Range(1, dst_gold.cols)), dst(cv::Range(1, dst_gold.rows), cv::Range(1, dst_gold.cols)), src.depth() == CV_32F ? 1e-4 : 1.0);
}
INSTANTIATE_TEST_CASE_P(ImgProc, PyrUp, Combine(
ALL_DEVICES,
Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_16UC1, CV_16UC3, CV_16UC4, CV_32FC1, CV_32FC3, CV_32FC4),
WHOLE_SUBMAT));
//////////////////////////////////////////////////////// ////////////////////////////////////////////////////////
// Canny // Canny
......
modules/gpu/test/test_pyramids.cpp 0 → 100644
View file @ 6397fa5b
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// Intel License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000, Intel Corporation, all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of Intel Corporation may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
#ifdef HAVE_CUDA
////////////////////////////////////////////////////////
// pyrDown
PARAM_TEST_CASE(PyrDown, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
{
cv::gpu::DeviceInfo devInfo;
cv::Size size;
int type;
bool useRoi;
virtual void SetUp()
{
devInfo = GET_PARAM(0);
size = GET_PARAM(1);
type = GET_PARAM(2);
useRoi = GET_PARAM(3);
cv::gpu::setDevice(devInfo.deviceID());
}
};
TEST_P(PyrDown, Accuracy)
{
cv::Mat src = randomMat(size, type);
cv::gpu::GpuMat dst = createMat(cv::Size((size.width + 1) / 2, (size.height + 1) / 2), type, useRoi);
cv::gpu::pyrDown(loadMat(src, useRoi), dst);
cv::Mat dst_gold;
cv::pyrDown(src, dst_gold);
EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-4 : 1.0);
}
INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrDown, testing::Combine(
ALL_DEVICES,
DIFFERENT_SIZES,
testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
WHOLE_SUBMAT));
////////////////////////////////////////////////////////
// pyrUp
PARAM_TEST_CASE(PyrUp, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
{
cv::gpu::DeviceInfo devInfo;
cv::Size size;
int type;
bool useRoi;
virtual void SetUp()
{
devInfo = GET_PARAM(0);
size = GET_PARAM(1);
type = GET_PARAM(2);
useRoi = GET_PARAM(3);
cv::gpu::setDevice(devInfo.deviceID());
}
};
TEST_P(PyrUp, Accuracy)
{
cv::Mat src = randomMat(size, type);
cv::gpu::GpuMat dst = createMat(cv::Size(size.width * 2, size.height * 2), type, useRoi);
cv::gpu::pyrUp(loadMat(src, useRoi), dst);
cv::Mat dst_gold;
cv::pyrUp(src, dst_gold);
EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-4 : 1.0);
}
INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrUp, testing::Combine(
ALL_DEVICES,
DIFFERENT_SIZES,
testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
WHOLE_SUBMAT));
#endif // HAVE_CUDA
modules/gpu/test/utility.hpp
View file @ 6397fa5b
...@@ -167,6 +167,8 @@ CV_FLAGS(DftFlags, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX ...@@ -167,6 +167,8 @@ CV_FLAGS(DftFlags, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX
#define DIFFERENT_SIZES testing::Values(cv::Size(128, 128), cv::Size(113, 113)) #define DIFFERENT_SIZES testing::Values(cv::Size(128, 128), cv::Size(113, 113))
#define WHOLE testing::Values(UseRoi(false))
#define SUBMAT testing::Values(UseRoi(true))
#define WHOLE_SUBMAT testing::Values(UseRoi(false), UseRoi(true)) #define WHOLE_SUBMAT testing::Values(UseRoi(false), UseRoi(true))
#define DIRECT_INVERSE testing::Values(Inverse(false), Inverse(true)) #define DIRECT_INVERSE testing::Values(Inverse(false), Inverse(true))
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment