In this work, a fast and robust method for MR brain segmentation is proposed. This method is based on a fast and robust fuzzy clustering algorithm that is initialized close to the searched solution in order to speed up the segmentation process. To validate the proposed method, we test it on some grayscale images and on a normal brain; brought from the BrainWeb Simulated Brain Database. The experimental results are important in both robustness to noise and running times standpoints.