MRI brain image segmentation by adaptive spatial deterministic annealing clustering

This paper describes a novel image segmentation algorithm suitable for MRI image segmentation. We introduce a new dissimilarity measure which incorporates the spatial connectivity. A fully automatic technique is developed to obtain the segmentation result and the new clustering objective function incorporates the spatial information. The weighting factor for neighborhood effect is adaptive to the image content. It enhances the smoothness towards piecewise-homogeneous region and reduces the edge-blurring effect. The experimental results with synthetic and MRI brain images demonstrate that the proposed method is effective in improving the segmentation and it outperforms the popular Markov random field algorithm