Image Clustering Based on Different Length Particle Swarm Optimization (DPSO)

Partitioning image pixels into several homogeneous regions is treated as the problem of clustering the pixels in the image matrix. This paper proposes an image clustering algorithm based on different length particle swarm optimization algorithm. Three evaluation criteria are used for the computation of the fitness of the particles of PSO based clustering algorithm. A novel Euclidean distance function is proposed based on the spatial and coordinate level distances of two image pixels towards measuring the similarity/dissimilarity. Different length particles are encoded in the PSO to minimize the user interaction with the program hence the execution time. PSO with different length particles automatically finds the number of cluster centers in the intensity space. The performance of the proposed algorithm is demonstrated by clustering different standard digital images. Results are compared with some well known existing algorithms.