In this paper, we propose a method to extract brain region from magnetic resonance imaging (MRI). The technique is commonly applied to assess brain atrophy acquired by taking MRI. The collected information is essential monitoring disease progress. Therefore it plays an important role in developing a computer-aided diagnosis of vascular dementia. For brain atrophy measurement, main regions, which should be focused and extracted to calculate an atrophy ratio from MRI, are white matter (WM) and gray matter (GM) and region containing cerebrospinal fluid (CSF). The proposed method helps us segment WM and GM regions from T2-weighted MRI. It consists of three main steps of normalizing pre-processing, clustering and post-processing. The first pre-processing step standardizes and enhances the brightness of brain region using complement image and Otsu’s method. The second step is clustering process using anisotropic diffusion filter and k-means clustering to separate brain region from background and non-bran region. Finally the post-processing step refines brain region extracted from the previous step using morphology technique. The experimental results on real T2-weighted MRI database demonstrate that our proposed approach has a good performance with high accuracy of 95.24%.