This paper presents a novel fully automated unsupervised framework for the brain tissue segmentation in magnetic resonance (MR) images. The framework is a combination of Bayesian-based adaptive mean shift, a priori spatial tissue probability maps and fuzzy c-means. Mean shift is employed to cluster the tissues in the joint spatial-intensity feature space and then a fuzzy c-means is applied with initialization by a priori spatial tissue probability maps to assign the clusters into three tissue types; white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF). The proposed framework is validated on a synthetic T1-weighted MR image with varying noise characteristics and spatial intensity inhomogeneity, obtained from the BrainWeb database as well as on 38 real T1-weighted MR images, obtained from the IBSR repository. The performance of the proposed framework is evaluated relative to the three widely used brain segmentation toolboxes: FAST, SPM and PVC, and the adaptive mean shift (AMS) and classical fuzzy c-means methods. The experimental results demonstrate the robustness of the proposed framework, and that it exhibits a higher degree of segmentation accuracy in segmenting both synthetic and real T1-weighted MR images compared to all competing methods.