Investigating particle swarm optimisation topologies for edge detection in noisy images

Abstract

This paper investigates the effects of applying different well-known static and dynamic neighbourhood topologies on the efficiency and effectiveness of a particle swarm optimisation-based edge detection algorithm. Our experiments show that the use of different topologies in a PSO-based edge detection algorithm does not have any significant effect on the accuracy of the algorithm for noisy images in most cases. That is in contrast to many reported results in the literature which claim that the selection of the neighbourhood topology affects the robustness of the algorithm to premature convergence and its accuracy. However, the fully connected topology in which all particles are connected to each other and exchange information performs more efficiently than other topologies in the PSO-based based edge detector. © 2011 Springer-Verlag.