Skip to Main Content
We implement a complete compressed domain video decoding pipeline that is optimized for both fast and memory efficient decoding. We combine fast DCT-domain inverse motion compensation algorithms with memory compression methods that take advantage of the sparse number of nonzero DCT coefficients within a coded block. We integrate these into an H.263 and MPEG-4 simple profile video decoder. We measure the performance of our video decoder along three dimensions: computational complexity, memory efficiency, and PSNR.