Frequency sampling design of 2-D quadrantally symmetric FIR filters by 2-D block decomposition technique

In this brief, two approaches-a direct and a progressive-for the frequency-sampling design of 2-D quadrantally symmetric FIR filters using efficient and recursive algorithms are presented. Allowing the samples to be distributed on sampling curves with shapes similar to the isocontours of the desired filter response, the method is suitable for the design of circular, elliptical, rectangular, or directional FIR filters. The method is based on a 2-D block decomposition technique and it is well suited to parallel computation, since almost all the operations are matrix multiplications and/or matrix-vector products. The proposed approaches can be also applied for computing 2-D inverse cosine transforms and can be extended to m-D design problems