Fast algorithm for generation of moment invariants

摘要

Moment invariants are important shape descriptors in computer vision. The method of decomposing the trigonometric function is suggested to obtain various moment invariants. Based on this method, the “multi-filter” algorithm is introduced as an efficient way to generate large numbers of moment invariants. A great deal of repeated computation on sub-polynomial is avoided. General explicit constructions of basic moment invariants are also provided. Furthermore, the proposed magnitude-normalized method makes invariants more stable and easier for classification.