Bounded Flexible Scale Mixture of Normal Distributions with Application to Image Segmentation

A bounded flexible scale mixture of normal (BFSMN) distributions is proposed as a novel device for modeling asymmetric and bounded data. Some characterizations and probabilistic properties of the BFSMN distributions and an extension to finite mixtures thereof are discussed. Based on a sort of selection mechanism, we design a feasible expectation-conditional maximization either algorithm to compute the maximum likelihood estimates of model parameters. To validate the effectiveness of the proposed methodology, we conduct experiments on both simulated data and real natural images and magnetic resonance images. The obtained results demonstrate the efficacy and usefulness of the proposed methodology.