Improvement on Image Edge Detection Using a Novel Variant of the Ant Colony System

Abstract

In this paper, an improved Ant Colony System algorithm applied to image edge detection is presented. During their movement on image, artificial ants establish pheromone graph which represents the image edge information. The ant movement is directed by the local variation of the image’s intensity values. To improve this method, supplementary behaviors are added to each ant in response to its local stimuli, i.e., the ant self-reproduces and directs its progenitors to an appropriate direction to explore more suitable areas. Moreover, it dies if it exceeds a specific iteration age and so the ineffective searches are eliminated. These additional behaviors allow diversifying the exploration performed by ants and also reinforcing the exploitation of these ants’ search experience. Proposed approach allows having more accurate and more complete edges. The performance is tested visually with various images and empirically with evaluation parameters.