Abstract

The multivariate Gaussian distribution function was frequently used as the component of the finite mixture model for image segmentation, however the clustering cannot be restricted to the normal distribution in the real data set. In order to make the cluster algorithm which is based on the mixture model have strong adaptability, a method of image segmentation using the elliptical contoured mixture model, which is constructed by the elliptical contoured distribution in the theorem of the generalized multivariate analysis, will be discussed in this paper. Moreover, the parameters of the model can be estimated according to the maximum likelihood and EM algorithm. And also, the pixels are classified the each regions under the rule of the maximum a posteriori. The experimental results with synthetic data and real image demonstrate the conclusion that the proposed method can get a better effectiveness and adaptability.

Paper Details