Abstract: This study presented a novel image registration algorithm for Magnet Resonance Image (MRI) using cascaded Pulse Coupled Neural Networks (PCNNs). Firstly, the two unregistrated images barycenters and foveation points are extracted respectively. Next, obtaining some clustering centers of the extracted foveations via fuzzy C-Means clustering (FCM) algorithm. Finally, using barycenter as coordinate origin and the corresponding clustering centers of foveations as characteristics to build coordinate system and attain the correlative registration parameters. The experimental results show that registration accuracy of the proposed algorithm is 95.7%. Meanwhile, it is robust and effective for MR image registration.