Gaussian Mixture Model with Markov Random Field for MR Image Segmentation
Huiguang He, Ke Lü, Bin Lv · 2006
In this paper, we propose a powerful fully automated classification method, which is based on Gaussian-mixture model with Markov random field (MRF). First, anisotropic diffusion is performed on the MR image to improve the signal noise ratio while keeping the edge; second, skull-stripping technique is applied to separate brain/non-brain tissue; third, histogram analysis is used to get the initial classification; finally, GMM-MRF is used to get the final classification. The method has been validated on simulated and real MR images for which gold standard segmentation is available. The experimental results show that the proposed method is more accurate and robust than currently available models.