What is the correct method for utilizing the gaussian filter function in the CUDA Npp library?

edit

The correct method for utilizing the gaussian filter function in the CUDA Npp library is as follows:

1. Include the necessary headers:

#include <npp.h>
#include <nppi_filtering_functions.h>

1. Allocate and initialize the NPP image and set the ROI:

NppiSize sizeROI = { width, height };
NppiSize maskSize = { maskWidth, maskHeight };
int nBufferSize = 0;
Npp8u* pBuffer = 0;
Npp8u* pSrcDst = 0;
int nSrcDstStep = 0;
NppiPoint anchor = { maskWidth / 2, maskHeight / 2 };

nppiMalloc(&pSrcDst, nSrcDstStep * height);
nppiSet_8u_C1R(0, pSrcDst, nSrcDstStep, sizeROI);

1. Calculate the size and allocate the buffer for the filter:

nppiFilterGaussBorderGetBufferSize(sizeROI, maskSize, &nBufferSize);
cudaMalloc(&pBuffer, nBufferSize);

1. Apply the filter on the image:

nppiFilterGaussBorder_8u_C1R(pSrcDst, nSrcDstStep, sizeROI, maskSize, anchor, pBuffer);

1. Free the memory:

cudaFree(pBuffer);
nppiFree(pSrcDst);

Note that the above code assumes that you have already defined the input and output size (width and height), mask size (maskWidth and maskHeight), and source/destination step (nSrcDstStep).