Analogic algorithm for point pattern matching with application to mammogram follow-up

Abstract

An analogic algorithm for registering control points in pairs of nonstructured texture images is described. The algorithm is particularly suitable in problems where precise pixel-wise registration is intractable, e.g., in registration of non-rigid objects. The algorithm is simulated on a digital computer as a sequence of cellular neural network (CNN) transients and logic operations. The proposed algorithm was applied to solve the correspondence problem in mammogram follow-up, the results of the performance evaluation of the algorithm are presented and discussed. We present a CNN realization of the proposed regional registration algorithm as an alternative to standard image processing implementation.