Genetic algorithms and grid technologies in clustering, an example: Clustering of images

Abstract

AbstractIn this paper we describe the development of an image retrieval system that is able to browse, cluster and classify large digital image databases. This work was motivated by the projects of the Visualization Centre of the Eötvös Loránd University, where such problems are to be solved. The system's functions are based on a Gaussian mixture model (GMM) representation of the images. Image matching is done by the distance measure of the representations, based on the approximation of the Kullback–Leibler divergence of the GMMs. The GMMs are estimated with an improved expectation maximization (EM) algorithm that avoids convergence to the boundary of the parameter space. These form the basis of the clustering, where a variant of a genetic algorithm is used. The suggested algorithm is able to work with a large number of images or objects, the grid technology is a useful tool for generating several runs simultaneously. Copyright © 2008 John Wiley & Sons, Ltd.