Image Segmentation based on Determinative Brain Storm optimization

Abstract

Brain Storm optimization (BSO) is a swarm-based, metaheuristic for global optimization, which has been inspired by the collective behavior of human beings. In this work, a novel BSO-based variant, Determinative BSO (DBSO), is proposed and applied for image segmentation. The proposed algorithm implements a cluster-merging strategy, inspired by the process of building a consensus among the members of a group with similar 'ideas'. It aims to prevent premature convergence and'jumping out' of local optima, in an optimization context for the determination of multiple image thresholds. Experiments on standard benchmark images are presented, which demonstrate that the proposed DBSO-based multilevel thresholding method obtains segmentation results of comparable or higher quality, in less iterations, than the ones obtained by state-of-the-art optimization-based multilevel thresholding methods. © 2020 IEEE.