Problem about gpu decoder

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m trying to learn how to use gpu decoder. I have read the sample program “cudaDecodeD3D9”. If I want to do some video postprocessing such as edge detection, how can I access the image data of each frame?

I have noticed the code below:

// perform post processing on the CUDA surface (performs colors space conversion and post processing)

				 // comment this out if we inclue the line of code seen above 

	cudaPostProcessFrame(&pDecodedFrame, nDecodedPitch, &pTextureData, nTexturePitch, g_pCudaModule->getModule(), gfpNV12toARGB,		   g_KernelSID);

I dont know how this code can perform colors space conversion. Which variable is the image data pointer?