Single photon emission computed tomography (SPECT) and magnetic resonance (MR) images are two common modalities in brain imaging society. Co-registration of brain SPECT and MR images has been used extensively in a number of clinical applications. The surface-based method and the mutual-information-based method are two major co-registration methods. The former aligns two volumetric images based on the cost function derived from brain surfaces, whereas the latter matches two volumetric images based on maximizing the mutual information (MI) between intensities of two images. To the best of our knowledge, these two methods have neither been compared, nor conjointly used to register the brain SPECT and MR images. Simulation results show that the surface-based method is more robust to the large deviation of initial guesses and computationally efficient, while the MI-based method provides better precision. The complementary features of these two methods motivated us to propose a hybrid strategy in which the surface-based method is employed to rapidly achieve a rough alignment followed by the MI-based method for fine tuning the results. Results from simulations suggested that only two iterations of MI calculation were required after surface-based registration and thereby lots of computation efforts can be saved.