Estimation of Multiple View Relations Considering Evolutionary Approaches

Abstract

Many computer vision problems require estimating geometric relations between two different views that contain a considerable number of unwanted abnormal data. Despite several robust techniques with different criteria have been proposed to solve such a problem, the Random Sampling Consensus (RANSAC) algorithm is by far the most well-known method. In this chapter, a method for robustly estimating multiple view relations from point correspondences is presented. The approach combines the RANSAC method and the Clonal Selection algorithm. Upon such combination, the method adopts a different sampling strategy in comparison to RANSAC in order to generate putative solutions. Under the new mechanism, new candidate solutions are iteratively built by considering the quality of models that have been generated by previous candidate solutions, rather than relying over a pure random selection as it is the case with classic RANSAC. The rules for the generation of candidate solutions (samples) are motivated by the behavior of the immunological system in human beings. As a result, the approach can substantially reduce the number of iterations but still preserves the robust capabilities of RANSAC. The method is generic and its use is illustrated by the estimation of fundamental matrices and homographies for synthetic and real images. Experimental results validate the efficiency of the resultant method in terms of accuracy, speed, and robustness.