Authors : Salima Ouadfel and Mohamed Batouche
Abstract: Image segmentation can be viewed as a labeling process in which image pixels are categorized into different classes that each class groups pixels sharing similar attributes. In this study, an unsupervised color image segmentation algorithm is presented using the Ant Colony System (ACS) algorithm and a Markov Random Field (MRF) segmentation model. The goal is to find a labeling, which is both piecewise smooth and consistent with the observed data image. We consider a energy function based on Markov Random Fields and we seek for the labeling that minimizes it. This is a combinatorial optimization problem difficult to solve. Ant algorithms are evolutionary algorithms where artificial ants are allowed to interact and cooperate to find the best solution to the given problem. The proposed algorithm is based on a population of simple agents, which construct candidate segmentation, by a relaxation labeling with respect to the contextual constraints. A local search is added to achieve further improvement. Experiments are performed on both synthetic and real images to compare the proposed algorithm to other metaheuristic techniques. The results show that our algorithm is quite competitive to other well known metaheuristic techniques.
Salima Ouadfel and Mohamed Batouche , 2006. Autonomous Agents-Based Approach for Color Image Segmentation. Asian Journal of Information Technology, 5: 528-535.