On the application of a fast polynomial transform and the Chinese remainder theorem to compute a two-dimensional convolution
T. K. Truong, I.S. Reed, Richard G. Lipes, Chang-Yu Wu · IEEE Transactions on Acoustics Speech and Signal Processing · 1981
A fast algorithm is developed to compute two dimensional convolutions of an array of d sub 1 X d sub 2 complex number points, where d sub 2 = 2(M) and d sub 1 = 2(m-r+) for some 1 or = r or = m. This algorithm requires fewer multiplications and about the same number of additions as the conventional fast fourier transform method for computing the two dimensional convolution. It also has the advantage that the operation of transposing the matrix of data can be avoided.