The invention discloses a brain tissue MRI image segmentation method based on fuzzy clustering and a Markov random field. The method is characterized by extracting gray-scale information of a low-frequency image obtained through multi-scale decomposition by utilizing a fuzzy clustering method, and carrying out image segmentation on the integrated gray-scale information through a single-point group potential energy function of a Markov random field model, thereby improving anti-interference capability to image noise. The method makes full use of the gray scale information of the image itself and space information between adjacent pixels, so that information can be used more fully, and the final image segmentation performance is allowed to be improved.