Block computation architectures for 2-D discrete wavelet transforms

In this paper, we will introduce two coarse-grained architectures for the block computations of 2-D separable DWT. Parallel-sequential architecture shows an improvement in computational complexity by a factor of JK, where J is the octave requirement and K is the filter size, when compared to sequential architecture. The parallel-parallel architecture has a computational complexity of O(N). Both architectures are scalable for different filter sizes and octave levels.