A novel modularized fast polynomial transform algorithm for two- dimensional convolutions (Q1190157)

The authors develop the ideas of \textit{H. J. Nussbaumer} [Digital filtering using polynomial transforms. Electron Letters 13, 386-387 (1977)] and of \textit{I. Pitas} and \textit{M. G. Strintzis} [IEEE Trans. Acoust. Speech Signal Process. ASSP-35, No. 3, 384-390 (1984)] to construct some polynomial transform algorithms for the corresponding two- dimensional (2-D) convolution. Earlier the so-called multi-D cyclic convolution algorithm was proposed, based on the factorization of multi- variable polynomial rings. However, the shortcoming of length variation still exists in this approach. The authors combine the above mentioned techniques to derive a novel modularized algorithm for 2-D convolution. A concrete example is given to show how the proposed algorithm works.