A PSO-based Algorithm with Subswarm Using Entropy and Uniformity for Image Segmentation

In the image segmentation field, it needs several iterations to find optimization thresholds or cluster center to segment images. in this paper, we embedded a scheme based on maximum entropy and uniformity into the particle swarm optimization with sub swarm structure named MEUPSOS to find the optimization threshold values iteratively. Instead of using the conventional PSO, swarm was divided into several sub swarms for the purpose of getting local optimal solutions with a fitness function based on maximal entropy and uniformity. Additionally, just one-swarm particles were used to replace k-swarm (k is the number of threshold values) particles in order to upgrade the computation performance. Then, the local optimal solutions ware used to update global parameter in the global swarm. through iterations updating the velocities and locations of particles, we can calculate the near optimal threshold values on an image based on the fitness functions of maximum entropy and uniformity. Finally, we can find that the proposed method can get more promising results than the other method.