A novel approach using computational fluid dynamics (CFD) and magnetic resonance image (MRI) is applied to model the auto-regulation and blood flow in the human brain. To provide a basic understanding of the auto-regulation mechanism in the brain, an anatomical Circle of Willis configuration is reconstructed from subject-specific magnetic resonance images using image segmentation methods and grid generation techniques. The three-dimensional unsteady incompressible Navier-Stokes equations are solved iteratively using the pseudocompressibility method and dual time stepping method. For the efficient simulation of three-dimensional time-dependent flows, parallel computations based on a domain decomposition method are performed. A simple auto-regulation algorithm is presented to model the dynamic peripheral resistance due to arteriolar contraction and dilatation. The present numerical methods are then used to simulate the auto-regulation of blood flow in the realistic Circle of Willis model with geometrical variants. The computed results show the correlation between abnormal vascular structures and the auto-regulation mechanism in the cerebral circulation.