Edge Detection Technique using Binary Particle Swarm Optimization

Edge detection is long established in computer eyesight applications such as article identification, shape matching, medical image classification, etc. For this reason, many edge detectors like LOG, Prewitt, Canny, etc. have been developed in the past in order to boost the grouping correctness of edge pixels. All these approaches work fine on images having minimum variation in intensity, however, their performance is not consistent on images having high-intensity variation. Therefore, in this paper "Binary Particle Swarm Optimization (BPSO)" based edge detection methodology minimizing multi-objective fitness function is proposed. Multi-objective fitness function is formulated by considering the weighted sum of five cost factors and all these cost factors are associated with four techniques of edge validation. The proposed approach is examined on 500 "BSD" images and results are compared with classical edge detectors (Canny, Prewitt) as well as computational intelligent techniques (ACO, GA) using the F score performance parameter. Performance of the proposed approach are consistent on all testing images and outperform all classical edge detectors, ACO and GA having average F score 0.2901 and have little standard deviation (0.0401). (C) 2020 The Authors. Published by Elsevier B.V.