New Attempts in Binary Image Registration

Abstract

The work reported in the paper addresses the problem of binary image registration. The proposed methodology is based on an evolution strategy (ES) developed to align two binary images. The fitness function is defined in terms of mutual information computed between the sensed image and the target image. We use various recombination schemes involving standard local/global convex and discrete crossover procedures as well as hybrid methods. The proposed hybridizations are developed based on some inner bond between the chromosome alleles. In the final stage we evaluate the similarity between the resulted image and the target using both quantitative and qualitative measures. Also, in order to experimentally establish the performances of our method, we compare it against one of the most commonly used procedures for image registration in case of rigid type perturbation, namely PAT registration. The experimental results together with some conclusive remarks regarding the quality of the proposed methodology are reported in the final part of the paper.