A modified gaussian mixture model algorithm for image segmentation based on dependable spatial constraints

In this paper, a modified Gaussian Mixture Model algorithm based on dependable spatial constraints is proposed for image segmentation. In order to enhance the segmentation performance, the proposed algorithm utilizes the consistence between the pixel and its local window to discriminate uncorrupted pixels from corrupted pixels. Then, using these uncorrupted pixels, the dependable spatial constraints are applied to influence the labeling of the pixel. In this way, the spatial information with high reliability is incorporated into the segmentation process, as a result, the segmentation accuracy is guaranteed to a great extent. The extensive segmentation experiments on synthetic, real and medical images demonstrate the effectiveness of the proposed algorithm.