Registration of successive DSA images using a simple genetic algorithm with a stochastic performance function

Abstract

The authors present a method aimed at automatic registration of successive images in a sequence of digital subtraction angiographies (DSA). It is based on the search for polynomial warping which minimizes a L/sub 1/ distance between the two images to be registered. The optimization procedure is carried out by a method based on a traditional genetic algorithm with a stochastic fitness function which operates only on randomly selected individuals. Application to the registration of successive DSA images has been made by considering a map of sites in the neighborhoods of contours. Good preliminary results have been obtained in a reasonable processing time on a sequential computer.<>