Two-dimensional DCT/DST universal computational structure for 2m * 2n block sizes

A refined generalized signal flow graph for the direct two-dimensional discrete cosine transform (2-D DCT) and discrete sine transform (2-D DST) computation (the so-called 2-D DCT/DST universal computational structure) is described. It represents a generalized unified approach to the fast 2-D DCT and 2-D DST computation for any 2 m * 2n block sizes, i.e., both square and rectangular blocks, including the one-dimensional (1-D) case. The regular structure, moderate arithmetic complexity, numerical stability, and multiple block size capability makes it suitable for VLSI or parallel implementation.