Rotation-invariant pattern recognition using circular harmonics and synthetic discriminant functions

Abstract

The use of circular harmonics to design correlation filters represents one of the most powerful approach in one-target recognition with in-plane rotation invariance. Recognition of more than one target can be achieved, on the other hand, using synthetic discriminant functions. In this paper we describe how can be combined these two techniques to design a filter which ensures rotation invariant recognition of multiple targets in parallel. The usefulness of the filter is demonstrated, for the case of two targets, by experimental results obtained from computer simulations.