A Novel Image Edge Detection Method Based on Multi-Population Ant Colony Optimization

Edge detection is a fundamental problem in image processing and computer vision. Recently ant colony optimization (ACO) algorithm has been used in edge detection, most existing ACO-based edge detection methods used single ant population and a fixed number of neighborhood pixels to calculate the gradient in the heuristic information for each pixel in transition probability. Those make the algorithm tend to obtain local optima. To enhance the accuracy of the ACO-based edge detection methods, a multiple-population strategy is utilized in this paper. The artificial ants are divided into two populations to obtain both advantages of global search and local search, one trends to make the ants move around optima to focus on local search, the other population make the ants move sharply from current position to jump out local optima and explore global optima. The experimental results show the effectiveness of the proposed method.