A two-dimensional discrete convolution algorithm
B. Arambepola · IEE Proceedings F Communications Radar and Signal Processing · 1984
A theoretical result concerning the discrete Fourier transform is derived and used to develop a transform algorithm for computing two-dimensional convolutions. The use of this algorithm minimises the number of arithmetic operations and the memory requirements in computing a convolution of order (M×N) using a transform processor or program designed for a length No, where M≤N0 and N≤N0.It is particularly suitable for computing convolutions whose orders are not powers of two using conventional fast Fourier transform processors. Methods of implementing the algorithm are also presented.