Fast recursive computation of local axial moments by using primitive kernel functions

Computation of local axial moments from images in a running mode (per pixel) is computationally very extensive procedure, especially, when the local moments must be computed for many orientation angles and far a large window size at the same time. This paper considers efficient computation of local axial moments by using the concept of primitive kernel functions. It allows to reduce the computation time from O(L-2) to O(1) per pixel, where LxL, is the window size.