Swarm Intelligence and Image Segmentation

Image segmentation plays an essential role in the interpretation of various kinds of images. Image segmentation techniques can be grouped into several categories such as edge-based segmentation, region-oriented segmentation, histogram thresholding, and clustering algorithms (Gonzalez & Woods, 1992). The aim of a clustering algorithm is to aggregate data into groups such that the data in each group share similar features while the data clusters are being distinct from each other. The K-means algorithm is a widely used method used for finding the structure of data (Tou & Gonzalez 1974). This unsupervised clustering technique has a strong tendency to get stuck into local minima when finding an optimal solution. Therefore, clustering results are heavily dependent on the initial cluster centers distribution. Hence, the search for good initial parameters is a challenging issue and the clustering algorithms require a great deal of experimentation to determine the input parameters for the optimal or suboptimal clustering results. Competitive learning model introduced in (Rumelhart & Zipser, 1986) is an interesting and powerful learning algorithm which can be used in unsupervised training for image classification (Hung, 1993). Simple Competitive Learning (SCL), shows stability over different run trials but this stable result is not always the global optima. In fact, in some cases the SCL converges to local optima over all run trials and the learning rate needs to be adjusted in the course of experimentation so that the global optimization can be achieved. There are a number of techniques, developed for optimization, inspired by the behaviours of natural systems (Pham & Karaboga, 2000). Swarm intelligence (SI) including Ant Colony Optimization (ACO) introduced in (Dorigo et al., 1996) and Particle Swarm Optimization (PSO) introduced in (Kennedy & Eberhart, 1995) has been introduced in the literature as an optimization technique. There are several SI approaches for data clustering in the literature which use clustering techniques such as K-means algorithm. In most of these approaches ACO or PSO are used to obtain the initial cluster centers for the K-means algorithm. We propose a hybrid algorithm which combines SI with K-means. We also use the same method to combine SI with SCL. Our aim is to make segmentation results of both K-means and SCL less dependent on the initial cluster centers and learning rate respectively. Hence, their results are more accurate and stabilized by employing the ACO and PSO optimization techniques. This improvement is due to the larger search space provided by these techniques. In addition, our