A spatial random-meaningful neighbourhood topology in pso for edge detection in noisy images

The continuity of edges is very important in some image processing applications but the detection of continuous edges is a very hard problem and is particularly time consuming in noisy images. The Canonical Particle Swarm Optimisation (CanPSO) has been used for the detection of continuous edges in such images. The Fully Informed Particle Swarm (FIPS) is another well-known version of PSO with interesting features to overcome noise but it has never been used to detect edges in noisy images. In this paper, the performance of CanPSO and FIPS is investigated for detecting edges in noisy images when they utilise different well-known static and dynamic topologies. A novel spatial random-meaningful topology is also developed and utilised within the PSO-based edge detection algorithm. Experimental results indicate that the localisation accuracy of the PSO-based edge detector with the novel topology is higher than other static and dynamic topologies in most cases.