Color image segmentation using adaptive mean shift and statistical model-based methods

Abstract

In this paper, we propose an unsupervised segmentation algorithm for color images based on Gaussian mixture models (GMMs). The number of mixture components is determined automatically by adaptive mean shift, in which local clusters are estimated by repeatedly searching for higher density points in feature vector space. For the estimation of parameters of GMMs, the mean field annealing expectation-maximization (EM) is employed. The mean field annealing EM provides a global optimal solution to overcome the local maxima problem in a mixture model. By combining the adaptive mean shift and the mean field annealing EM, natural color images are segmented automatically without over-segmentation or isolated regions. The experiments show that the proposed algorithm can produce satisfactory segmentation without any a priori information.