Skip to content
Snippets Groups Projects
Commit 8728ba30 authored by Eric Wait's avatar Eric Wait
Browse files

Contrast enhancement now works faster by not using biggest overlap for each kernel

parent 88ccef47
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
src/c/Common/CudaContrastEnhancement.cuh
+ 111
− 49
View file @ 8728ba30
......@@ -14,53 +14,115 @@ template <class PixelType>
PixelType* contrastEnhancement(const PixelType* imageIn, Vec<size_t> dims, Vec<float> sigmas, Vec<size_t> neighborhood,
PixelType** imageOut=NULL, int device=0)
{
PixelType* imOut = setUpOutIm(dims, imageOut);
PixelType minVal = std::numeric_limits<PixelType>::lowest();
PixelType maxVal = std::numeric_limits<PixelType>::max();
neighborhood = neighborhood.clamp(Vec<size_t>(1,1,1),dims);
float* hostKernel;
Vec<int> gaussIterations(0,0,0);
Vec<size_t> sizeconstKernelDims = createGaussianKernel(sigmas,&hostKernel,gaussIterations);
HANDLE_ERROR(cudaMemcpyToSymbol(cudaConstKernel, hostKernel, sizeof(float)*
(sizeconstKernelDims.x+sizeconstKernelDims.y+sizeconstKernelDims.z)));
cudaDeviceProp props;
cudaGetDeviceProperties(&props,device);
size_t availMem, total;
cudaMemGetInfo(&availMem,&total);
std::vector<ImageChunk> chunks = calculateBuffers<double>(dims,3,(size_t)(availMem*MAX_MEM_AVAIL),props,sizeconstKernelDims);
Vec<size_t> maxDeviceDims;
setMaxDeviceDims(chunks, maxDeviceDims);
CudaDeviceImages<double> deviceImages(3,maxDeviceDims,device);
for (std::vector<ImageChunk>::iterator curChunk=chunks.begin(); curChunk!=chunks.end(); ++curChunk)
{
deviceImages.setAllDims(curChunk->getFullChunkSize());
curChunk->sendROI(imageIn,dims,deviceImages.getCurBuffer());
runGaussIterations(gaussIterations, curChunk, deviceImages, sizeconstKernelDims,device);
curChunk->sendROI(imageIn,dims,deviceImages.getNextBuffer());
cudaAddTwoImagesWithFactor<double><<<curChunk->blocks,curChunk->threads>>>(*(deviceImages.getNextBuffer()),*(deviceImages.getCurBuffer()),
*(deviceImages.getThirdBuffer()),-1.0,minVal,maxVal);
DEBUG_KERNEL_CHECK();
deviceImages.setNthBuffCurent(3);
runMedianFilter(props, curChunk, neighborhood, deviceImages);
curChunk->retriveROI(imOut,dims,deviceImages.getCurBuffer());
}
return imOut;
PixelType* imGauss = gaussianFilter<PixelType>(imageIn,dims,sigmas,NULL,device);
PixelType* imSub = addImageWith<PixelType>(imageIn,imGauss,dims,-1.0,NULL,device);
delete[] imGauss;
return medianFilter(imSub,dims,neighborhood,imageOut,device);
// PixelType* imOut = setUpOutIm(dims, imageOut);
//
// PixelType minVal = std::numeric_limits<PixelType>::lowest();
// PixelType maxVal = std::numeric_limits<PixelType>::max();
//
// neighborhood = neighborhood.clamp(Vec<size_t>(1,1,1),dims);
//
// float* hostKernel;
//
// Vec<int> gaussIterations(0,0,0);
// Vec<size_t> sizeconstKernelDims = createGaussianKernel(sigmas,&hostKernel,gaussIterations);
// HANDLE_ERROR(cudaMemcpyToSymbol(cudaConstKernel, hostKernel, sizeof(float)*
// (sizeconstKernelDims.x+sizeconstKernelDims.y+sizeconstKernelDims.z)));
//
// cudaDeviceProp props;
// cudaGetDeviceProperties(&props,device);
//
// size_t availMem, total;
// cudaMemGetInfo(&availMem,&total);
//
// std::vector<ImageChunk> chunks = calculateBuffers<float>(dims,3,(size_t)(availMem*MAX_MEM_AVAIL),props,sizeconstKernelDims);
//
// Vec<size_t> maxDeviceDims;
// setMaxDeviceDims(chunks, maxDeviceDims);
//
// CudaDeviceImages<float> deviceImages(3,maxDeviceDims,device);
//
// for (std::vector<ImageChunk>::iterator curChunk=chunks.begin(); curChunk!=chunks.end(); ++curChunk)
// {
// deviceImages.setAllDims(curChunk->getFullChunkSize());
//
// curChunk->sendROI(imageIn,dims,deviceImages.getCurBuffer());
//
// runGaussIterations(gaussIterations, curChunk, deviceImages, sizeconstKernelDims,device);
//
// curChunk->sendROI(imageIn,dims,deviceImages.getNextBuffer());
//
// cudaAddTwoImagesWithFactor<float><<<curChunk->blocks,curChunk->threads>>>(*(deviceImages.getNextBuffer()),*(deviceImages.getCurBuffer()),
// *(deviceImages.getThirdBuffer()),-1.0,(float)minVal,(float)maxVal);
// DEBUG_KERNEL_CHECK();
//
// deviceImages.setNthBuffCurent(3);
//
// runMedianFilter<float>(props, curChunk, neighborhood, deviceImages);
//
// curChunk->retriveROI<float,PixelType>(imOut,dims,deviceImages.getCurBuffer());
// }
//return imOut;
}
//double* contrastEnhancement(const double* imageIn, Vec<size_t> dims, Vec<float> sigmas, Vec<size_t> neighborhood,
// double** imageOut=NULL, int device=0)
//{
// double* imOut = setUpOutIm(dims, imageOut);
//
// double minVal = std::numeric_limits<double>::lowest();
// double maxVal = std::numeric_limits<double>::max();
//
// neighborhood = neighborhood.clamp(Vec<size_t>(1,1,1),dims);
//
// float* hostKernel;
//
// Vec<int> gaussIterations(0,0,0);
// Vec<size_t> sizeconstKernelDims = createGaussianKernel(sigmas,&hostKernel,gaussIterations);
// HANDLE_ERROR(cudaMemcpyToSymbol(cudaConstKernel, hostKernel, sizeof(float)*
// (sizeconstKernelDims.x+sizeconstKernelDims.y+sizeconstKernelDims.z)));
//
// cudaDeviceProp props;
// cudaGetDeviceProperties(&props,device);
//
// size_t availMem, total;
// cudaMemGetInfo(&availMem,&total);
//
// std::vector<ImageChunk> chunks = calculateBuffers<double>(dims,3,(size_t)(availMem*MAX_MEM_AVAIL),props,sizeconstKernelDims);
//
// Vec<size_t> maxDeviceDims;
// setMaxDeviceDims(chunks, maxDeviceDims);
//
// CudaDeviceImages<double> deviceImages(3,maxDeviceDims,device);
//
// for (std::vector<ImageChunk>::iterator curChunk=chunks.begin(); curChunk!=chunks.end(); ++curChunk)
// {
// deviceImages.setAllDims(curChunk->getFullChunkSize());
//
// curChunk->sendROI(imageIn,dims,deviceImages.getCurBuffer());
//
// runGaussIterations(gaussIterations, curChunk, deviceImages, sizeconstKernelDims,device);
//
// curChunk->sendROI(imageIn,dims,deviceImages.getNextBuffer());
//
// cudaAddTwoImagesWithFactor<double><<<curChunk->blocks,curChunk->threads>>>(*(deviceImages.getNextBuffer()),*(deviceImages.getCurBuffer()),
// *(deviceImages.getThirdBuffer()),-1.0,minVal,maxVal);
// DEBUG_KERNEL_CHECK();
//
// deviceImages.setNthBuffCurent(3);
//
// runMedianFilter(props, curChunk, neighborhood, deviceImages);
//
// curChunk->retriveROI(imOut,dims,deviceImages.getCurBuffer());
// }
//
// return imOut;
//}
\ No newline at end of file
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