Skip to content

O
opencv
Commit e1e5047b authored by Vladislav Vinogradov's avatar Vladislav Vinogradov
Browse files
Options

added gpu::LUT for CV_8UC3 type, added gpu::cvtColor for BGR2BGR5x5, minor fix in tests.

parent 1b8c0000
Expand all Show whitespace changes
Inline Side-by-side
Showing with 85 additions and 93 deletions
modules/gpu/src/arithm.cpp
View file @ e1e5047b
...@@ -266,13 +266,13 @@ double cv::gpu::norm(const GpuMat& src1, const GpuMat& src2, int normType) ...@@ -266,13 +266,13 @@ double cv::gpu::norm(const GpuMat& src1, const GpuMat& src2, int normType)
sz.height = src1.rows; sz.height = src1.rows;
int funcIdx = normType >> 1; int funcIdx = normType >> 1;
Scalar retVal; double retVal;
nppSafeCall( npp_norm_diff_func[funcIdx](src1.ptr<Npp8u>(), src1.step, nppSafeCall( npp_norm_diff_func[funcIdx](src1.ptr<Npp8u>(), src1.step,
src2.ptr<Npp8u>(), src2.step, src2.ptr<Npp8u>(), src2.step,
sz, retVal.val) ); sz, &retVal) );
return retVal[0]; return retVal;
} }
//////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////
...@@ -309,9 +309,6 @@ Scalar cv::gpu::sum(const GpuMat& src) ...@@ -309,9 +309,6 @@ Scalar cv::gpu::sum(const GpuMat& src)
{ {
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4); CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
NppiSize sz; NppiSize sz;
sz.width = src.cols; sz.width = src.cols;
sz.height = src.rows; sz.height = src.rows;
...@@ -336,8 +333,6 @@ Scalar cv::gpu::sum(const GpuMat& src) ...@@ -336,8 +333,6 @@ Scalar cv::gpu::sum(const GpuMat& src)
nppSafeCall( nppiSum_8u_C4R(src.ptr<Npp8u>(), src.step, sz, buf.ptr<Npp32s>(), res.val) ); nppSafeCall( nppiSum_8u_C4R(src.ptr<Npp8u>(), src.step, sz, buf.ptr<Npp32s>(), res.val) );
return res; return res;
} }
} }
//////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////
...@@ -371,28 +366,54 @@ void cv::gpu::LUT(const GpuMat& src, const Mat& lut, GpuMat& dst) ...@@ -371,28 +366,54 @@ void cv::gpu::LUT(const GpuMat& src, const Mat& lut, GpuMat& dst)
{ {
public: public:
Npp32s pLevels[256]; Npp32s pLevels[256];
const Npp32s* pLevels3[3];
int nValues3[3];
LevelsInit() LevelsInit()
{ {
nValues3[0] = nValues3[1] = nValues3[2] = 256;
for (int i = 0; i < 256; ++i) for (int i = 0; i < 256; ++i)
pLevels[i] = i; pLevels[i] = i;
pLevels3[0] = pLevels3[1] = pLevels3[2] = pLevels;
} }
}; };
static LevelsInit lvls; static LevelsInit lvls;
int cn = src.channels(); int cn = src.channels();
CV_Assert(src.type() == CV_8UC1); CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC3);
CV_Assert(lut.depth() == CV_32SC1 && lut.rows * lut.cols == 256 && lut.isContinuous()); CV_Assert(lut.depth() == CV_8U && (lut.channels() == 1 || lut.channels() == cn) && lut.rows * lut.cols == 256 && lut.isContinuous());
dst.create(src.size(), src.type()); dst.create(src.size(), CV_MAKETYPE(lut.depth(), cn));
NppiSize sz; NppiSize sz;
sz.height = src.rows; sz.height = src.rows;
sz.width = src.cols; sz.width = src.cols;
Mat nppLut;
lut.convertTo(nppLut, CV_32S);
if (src.type() == CV_8UC1)
{
nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), src.step, dst.ptr<Npp8u>(), dst.step, sz, nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), src.step, dst.ptr<Npp8u>(), dst.step, sz,
lut.ptr<Npp32s>(), lvls.pLevels, 256) ); nppLut.ptr<Npp32s>(), lvls.pLevels, 256) );
}
else
{
Mat nppLut3[3];
const Npp32s* pValues3[3];
if (nppLut.channels() == 1)
pValues3[0] = pValues3[1] = pValues3[2] = nppLut.ptr<Npp32s>();
else
{
cv::split(nppLut, nppLut3);
pValues3[0] = nppLut3[0].ptr<Npp32s>();
pValues3[1] = nppLut3[1].ptr<Npp32s>();
pValues3[2] = nppLut3[2].ptr<Npp32s>();
}
nppSafeCall( nppiLUT_Linear_8u_C3R(src.ptr<Npp8u>(), src.step, dst.ptr<Npp8u>(), dst.step, sz,
pValues3, lvls.pLevels3, lvls.nValues3) );
}
} }
#endif /* !defined (HAVE_CUDA) */ #endif /* !defined (HAVE_CUDA) */
\ No newline at end of file
modules/gpu/src/cuda/color.cu
View file @ e1e5047b
This diff is collapsed.
modules/gpu/src/imgproc_gpu.cpp
View file @ e1e5047b
...@@ -81,13 +81,16 @@ namespace cv { namespace gpu ...@@ -81,13 +81,16 @@ namespace cv { namespace gpu
void reprojectImageTo3D_gpu(const DevMem2D& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream); void reprojectImageTo3D_gpu(const DevMem2D& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
void reprojectImageTo3D_gpu(const DevMem2D_<short>& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream); void reprojectImageTo3D_gpu(const DevMem2D_<short>& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
void swapChannels_gpu(const DevMem2D& src, const DevMem2D& dst, int cn, const int* coeffs, cudaStream_t stream); void swapChannels_gpu_8u(const DevMem2D& src, const DevMem2D& dst, int cn, const int* coeffs, cudaStream_t stream);
void swapChannels_gpu(const DevMem2D_<ushort>& src, const DevMem2D_<ushort>& dst, int cn, const int* coeffs, cudaStream_t stream); void swapChannels_gpu_16u(const DevMem2D& src, const DevMem2D& dst, int cn, const int* coeffs, cudaStream_t stream);
void swapChannels_gpu(const DevMem2Df& src, const DevMem2Df& dst, int cn, const int* coeffs, cudaStream_t stream); void swapChannels_gpu_32f(const DevMem2D& src, const DevMem2D& dst, int cn, const int* coeffs, cudaStream_t stream);
void RGB2RGB_gpu(const DevMem2D& src, int srccn, const DevMem2D& dst, int dstcn, int bidx, cudaStream_t stream); void RGB2RGB_gpu_8u(const DevMem2D& src, int srccn, const DevMem2D& dst, int dstcn, int bidx, cudaStream_t stream);
void RGB2RGB_gpu(const DevMem2D_<ushort>& src, int srccn, const DevMem2D_<ushort>& dst, int dstcn, int bidx, cudaStream_t stream); void RGB2RGB_gpu_16u(const DevMem2D& src, int srccn, const DevMem2D& dst, int dstcn, int bidx, cudaStream_t stream);
void RGB2RGB_gpu(const DevMem2Df& src, int srccn, const DevMem2Df& dst, int dstcn, int bidx, cudaStream_t stream); void RGB2RGB_gpu_32f(const DevMem2D& src, int srccn, const DevMem2D& dst, int dstcn, int bidx, cudaStream_t stream);
void RGB5x52RGB_gpu(const DevMem2D& src, int green_bits, const DevMem2D& dst, int dstcn, int bidx, cudaStream_t stream);
void RGB2RGB5x5_gpu(const DevMem2D& src, int srccn, const DevMem2D& dst, int green_bits, int bidx, cudaStream_t stream);
void Gray2RGB_gpu(const DevMem2D& src, const DevMem2D& dst, int dstcn, cudaStream_t stream); void Gray2RGB_gpu(const DevMem2D& src, const DevMem2D& dst, int dstcn, cudaStream_t stream);
void Gray2RGB_gpu(const DevMem2D_<ushort>& src, const DevMem2D_<ushort>& dst, int dstcn, cudaStream_t stream); void Gray2RGB_gpu(const DevMem2D_<ushort>& src, const DevMem2D_<ushort>& dst, int dstcn, cudaStream_t stream);
...@@ -245,38 +248,36 @@ namespace ...@@ -245,38 +248,36 @@ namespace
out.create(sz, CV_MAKETYPE(depth, dcn)); out.create(sz, CV_MAKETYPE(depth, dcn));
if( depth == CV_8U ) if( depth == CV_8U )
improc::RGB2RGB_gpu((DevMem2D)src, scn, (DevMem2D)out, dcn, bidx, stream); improc::RGB2RGB_gpu_8u(src, scn, out, dcn, bidx, stream);
else if( depth == CV_16U ) else if( depth == CV_16U )
improc::RGB2RGB_gpu((DevMem2D_<unsigned short>)src, scn, (DevMem2D_<unsigned short>)out, dcn, bidx, stream); improc::RGB2RGB_gpu_16u(src, scn, out, dcn, bidx, stream);
else else
improc::RGB2RGB_gpu((DevMem2Df)src, scn, (DevMem2Df)out, dcn, bidx, stream); improc::RGB2RGB_gpu_32f(src, scn, out, dcn, bidx, stream);
break; break;
//case CV_BGR2BGR565: case CV_BGR2BGR555: case CV_RGB2BGR565: case CV_RGB2BGR555: case CV_BGR2BGR565: case CV_BGR2BGR555: case CV_RGB2BGR565: case CV_RGB2BGR555:
//case CV_BGRA2BGR565: case CV_BGRA2BGR555: case CV_RGBA2BGR565: case CV_RGBA2BGR555: case CV_BGRA2BGR565: case CV_BGRA2BGR555: case CV_RGBA2BGR565: case CV_RGBA2BGR555:
// CV_Assert( (scn == 3 || scn == 4) && depth == CV_8U ); CV_Assert( (scn == 3 || scn == 4) && depth == CV_8U );
// dst.create(sz, CV_8UC2); out.create(sz, CV_8UC2);
//
// CvtColorLoop(src, dst, RGB2RGB5x5(scn, improc::RGB2RGB5x5_gpu(src, scn, out, code == CV_BGR2BGR565 || code == CV_RGB2BGR565 ||
// code == CV_BGR2BGR565 || code == CV_BGR2BGR555 || code == CV_BGRA2BGR565 || code == CV_RGBA2BGR565 ? 6 : 5,
// code == CV_BGRA2BGR565 || code == CV_BGRA2BGR555 ? 0 : 2, code == CV_BGR2BGR565 || code == CV_BGR2BGR555 ||
// code == CV_BGR2BGR565 || code == CV_RGB2BGR565 || code == CV_BGRA2BGR565 || code == CV_BGRA2BGR555 ? 0 : 2,
// code == CV_BGRA2BGR565 || code == CV_RGBA2BGR565 ? 6 : 5 // green bits stream);
// )); break;
// break;
//case CV_BGR5652BGR: case CV_BGR5552BGR: case CV_BGR5652RGB: case CV_BGR5552RGB: //case CV_BGR5652BGR: case CV_BGR5552BGR: case CV_BGR5652RGB: case CV_BGR5552RGB:
//case CV_BGR5652BGRA: case CV_BGR5552BGRA: case CV_BGR5652RGBA: case CV_BGR5552RGBA: //case CV_BGR5652BGRA: case CV_BGR5552BGRA: case CV_BGR5652RGBA: case CV_BGR5552RGBA:
// if(dcn <= 0) dcn = 3; // if(dcn <= 0) dcn = 3;
// CV_Assert( (dcn == 3 || dcn == 4) && scn == 2 && depth == CV_8U ); // CV_Assert( (dcn == 3 || dcn == 4) && scn == 2 && depth == CV_8U );
// dst.create(sz, CV_MAKETYPE(depth, dcn)); // out.create(sz, CV_MAKETYPE(depth, dcn));
//
// CvtColorLoop(src, dst, RGB5x52RGB(dcn, // improc::RGB5x52RGB_gpu(src, code == CV_BGR2BGR565 || code == CV_RGB2BGR565 ||
// code == CV_BGR5652BGR || code == CV_BGR5552BGR || // code == CV_BGRA2BGR565 || code == CV_RGBA2BGR565 ? 6 : 5, out, dcn,
// code == CV_BGR5652BGRA || code == CV_BGR5552BGRA ? 0 : 2, // blue idx // code == CV_BGR2BGR565 || code == CV_BGR2BGR555 ||
// code == CV_BGR5652BGR || code == CV_BGR5652RGB || // code == CV_BGRA2BGR565 || code == CV_BGRA2BGR555 ? 0 : 2,
// code == CV_BGR5652BGRA || code == CV_BGR5652RGBA ? 6 : 5 // green bits // stream);
// ));
// break; // break;
case CV_BGR2GRAY: case CV_BGRA2GRAY: case CV_RGB2GRAY: case CV_RGBA2GRAY: case CV_BGR2GRAY: case CV_BGRA2GRAY: case CV_RGB2GRAY: case CV_RGBA2GRAY:
...@@ -329,7 +330,7 @@ namespace ...@@ -329,7 +330,7 @@ namespace
nppSafeCall( nppiRGBToYCbCr_8u_C3R(src.ptr<Npp8u>(), src.step, out.ptr<Npp8u>(), out.step, nppsz) ); nppSafeCall( nppiRGBToYCbCr_8u_C3R(src.ptr<Npp8u>(), src.step, out.ptr<Npp8u>(), out.step, nppsz) );
{ {
static int coeffs[] = {0, 2, 1}; static int coeffs[] = {0, 2, 1};
improc::swapChannels_gpu((DevMem2D)out, (DevMem2D)out, 3, coeffs, 0); improc::swapChannels_gpu_8u(out, out, 3, coeffs, 0);
} }
break; break;
...@@ -341,7 +342,7 @@ namespace ...@@ -341,7 +342,7 @@ namespace
{ {
static int coeffs[] = {0, 2, 1}; static int coeffs[] = {0, 2, 1};
GpuMat src1(src.size(), src.type()); GpuMat src1(src.size(), src.type());
improc::swapChannels_gpu((DevMem2D)src, (DevMem2D)src1, 3, coeffs, 0); improc::swapChannels_gpu_8u(src, src1, 3, coeffs, 0);
nppSafeCall( nppiYCbCrToRGB_8u_C3R(src1.ptr<Npp8u>(), src1.step, out.ptr<Npp8u>(), out.step, nppsz) ); nppSafeCall( nppiYCbCrToRGB_8u_C3R(src1.ptr<Npp8u>(), src1.step, out.ptr<Npp8u>(), out.step, nppsz) );
} }
break; break;
......
tests/gpu/src/arithm.cpp
View file @ e1e5047b
...@@ -131,45 +131,6 @@ void CV_GpuArithmTest::run( int ) ...@@ -131,45 +131,6 @@ void CV_GpuArithmTest::run( int )
testResult = CvTS::FAIL_MISMATCH; testResult = CvTS::FAIL_MISMATCH;
} }
} }
///!!! author, please remove commented code if loop above is equivalent.
/*ts->printf(CvTS::LOG, "\n========Start test 8UC1========\n");
if (test(CV_8UC1) == CvTS::OK)
ts->printf(CvTS::LOG, "\nSUCCESS\n");
else
{
ts->printf(CvTS::LOG, "\nFAIL\n");
testResult = CvTS::FAIL_GENERIC;
}
ts->printf(CvTS::LOG, "\n========Start test 8UC3========\n");
if (test(CV_8UC3) == CvTS::OK)
ts->printf(CvTS::LOG, "\nSUCCESS\n");
else
{
ts->printf(CvTS::LOG, "\nFAIL\n");
testResult = CvTS::FAIL_GENERIC;
}
ts->printf(CvTS::LOG, "\n========Start test 8UC4========\n");
if (test(CV_8UC4) == CvTS::OK)
ts->printf(CvTS::LOG, "\nSUCCESS\n");
else
{
ts->printf(CvTS::LOG, "\nFAIL\n");
testResult = CvTS::FAIL_GENERIC;
}
ts->printf(CvTS::LOG, "\n========Start test 32FC1========\n");
if (test(CV_32FC1) == CvTS::OK)
ts->printf(CvTS::LOG, "\nSUCCESS\n");
else
{
ts->printf(CvTS::LOG, "\nFAIL\n");
testResult = CvTS::FAIL_GENERIC;
}*/
} }
catch(const cv::Exception& e) catch(const cv::Exception& e)
{ {
...@@ -551,19 +512,18 @@ struct CV_GpuNppImageLUTTest : public CV_GpuArithmTest ...@@ -551,19 +512,18 @@ struct CV_GpuNppImageLUTTest : public CV_GpuArithmTest
int test( const Mat& mat1, const Mat& ) int test( const Mat& mat1, const Mat& )
{ {
if (mat1.type() != CV_8UC1) if (mat1.type() != CV_8UC1 && mat1.type() != CV_8UC3)
{ {
ts->printf(CvTS::LOG, "\nUnsupported type\n"); ts->printf(CvTS::LOG, "\nUnsupported type\n");
return CvTS::OK; return CvTS::OK;
} }
cv::Mat lut(1, 256, CV_32SC1); cv::Mat lut(1, 256, CV_8UC1);
cv::RNG rng(*ts->get_rng()); cv::RNG rng(*ts->get_rng());
rng.fill(lut, cv::RNG::UNIFORM, cv::Scalar::all(100), cv::Scalar::all(200)); rng.fill(lut, cv::RNG::UNIFORM, cv::Scalar::all(100), cv::Scalar::all(200));
cv::Mat cpuRes; cv::Mat cpuRes;
cv::LUT(mat1, lut, cpuRes); cv::LUT(mat1, lut, cpuRes);
cpuRes.convertTo(cpuRes, CV_8U);
cv::gpu::GpuMat gpuRes; cv::gpu::GpuMat gpuRes;
cv::gpu::LUT(GpuMat(mat1), lut, gpuRes); cv::gpu::LUT(GpuMat(mat1), lut, gpuRes);
......
tests/gpu/src/gputest_main.cpp
View file @ e1e5047b
...@@ -46,6 +46,18 @@ CvTS test_system; ...@@ -46,6 +46,18 @@ CvTS test_system;
const char* blacklist[] = const char* blacklist[] =
{ {
"GPU-NppImageSum", "GPU-NppImageSum",
"GPU-MatOperatorAsyncCall",
//"GPU-NppErode",
//"GPU-NppDilate",
//"GPU-NppMorphologyEx",
//"GPU-NppImageDivide",
//"GPU-NppImageMeanStdDev",
//"GPU-NppImageMinNax",
//"GPU-NppImageResize",
//"GPU-NppImageWarpAffine",
//"GPU-NppImageWarpPerspective",
//"GPU-NppImageIntegral",
//"GPU-NppImageBlur",
0 0
}; };
......
tests/gpu/src/imgproc_gpu.cpp
View file @ e1e5047b
...@@ -223,8 +223,8 @@ struct CV_GpuNppImageResizeTest : public CV_GpuImageProcTest ...@@ -223,8 +223,8 @@ struct CV_GpuNppImageResizeTest : public CV_GpuImageProcTest
return CvTS::OK; return CvTS::OK;
} }
int interpolations[] = {INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_LANCZOS4}; int interpolations[] = {INTER_NEAREST, INTER_LINEAR, /*INTER_CUBIC,*/ /*INTER_LANCZOS4*/};
const char* interpolations_str[] = {"INTER_NEAREST", "INTER_LINEAR", "INTER_CUBIC", "INTER_LANCZOS4"}; const char* interpolations_str[] = {"INTER_NEAREST", "INTER_LINEAR", /*"INTER_CUBIC",*/ /*"INTER_LANCZOS4"*/};
int interpolations_num = sizeof(interpolations) / sizeof(int); int interpolations_num = sizeof(interpolations) / sizeof(int);
int test_res = CvTS::OK; int test_res = CvTS::OK;
...@@ -443,9 +443,6 @@ struct CV_GpuNppImageBlurTest : public CV_GpuImageProcTest ...@@ -443,9 +443,6 @@ struct CV_GpuNppImageBlurTest : public CV_GpuImageProcTest
GpuMat gpudst; GpuMat gpudst;
cv::gpu::blur(gpu1, gpudst, Size(ksizes[i], ksizes[i])); cv::gpu::blur(gpu1, gpudst, Size(ksizes[i], ksizes[i]));
cv::Mat c;
cv::absdiff(cpudst, gpudst, c);
if (CheckNorm(cpudst, gpudst) != CvTS::OK) if (CheckNorm(cpudst, gpudst) != CvTS::OK)
test_res = CvTS::FAIL_GENERIC; test_res = CvTS::FAIL_GENERIC;
} }
...@@ -459,7 +456,7 @@ struct CV_GpuNppImageBlurTest : public CV_GpuImageProcTest ...@@ -459,7 +456,7 @@ struct CV_GpuNppImageBlurTest : public CV_GpuImageProcTest
class CV_GpuCvtColorTest : public CvTest class CV_GpuCvtColorTest : public CvTest
{ {
public: public:
CV_GpuCvtColorTest() : CvTest("GPU-NppCvtColor", "cvtColor") {} CV_GpuCvtColorTest() : CvTest("GPU-CvtColor", "cvtColor") {}
~CV_GpuCvtColorTest() {}; ~CV_GpuCvtColorTest() {};
protected: protected:
...@@ -501,8 +498,8 @@ void CV_GpuCvtColorTest::run( int ) ...@@ -501,8 +498,8 @@ void CV_GpuCvtColorTest::run( int )
try try
{ {
//run tests //run tests
int codes[] = {CV_BGR2RGB, CV_RGB2YCrCb, CV_YCrCb2RGB, CV_RGB2RGBA, CV_RGBA2BGRA, CV_BGRA2GRAY, CV_GRAY2RGB}; int codes[] = { CV_BGR2RGB, /* CV_RGB2YCrCb, CV_YCrCb2RGB,*/ CV_RGB2RGBA, CV_RGBA2BGRA, CV_BGRA2GRAY, CV_GRAY2RGB, CV_RGB2BGR555/*, CV_BGR5552BGR/*, CV_BGR2BGR565, CV_BGR5652RGB*/};
const char* codes_str[] = {"CV_BGR2RGB", "CV_RGB2YCrCb", "CV_YCrCb2RGB", "CV_RGB2RGBA", "CV_RGBA2BGRA", "CV_BGRA2GRAY", "CV_GRAY2RGB"}; const char* codes_str[] = {"CV_BGR2RGB", /*"CV_RGB2YCrCb", "CV_YCrCb2RGB",*/ "CV_RGB2RGBA", "CV_RGBA2BGRA", "CV_BGRA2GRAY", "CV_GRAY2RGB", "CV_RGB2BGR555"/*, "CV_BGR5552BGR"/*, "CV_BGR2BGR565", "CV_BGR5652RGB"*/};
int codes_num = sizeof(codes) / sizeof(int); int codes_num = sizeof(codes) / sizeof(int);
for (int i = 0; i < codes_num; ++i) for (int i = 0; i < codes_num; ++i)
......
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