This research proposes and implements a new fully automatic fast registration method for 2D Computed Tomography images. In this proposed method, the images are smoothened and edges are detected using Laplacian of Gaussian filter. Fast Fourier transform is applied to the source and reference images to split the phase and magnitude. The registered image is reconstructed using Inverse Fourier Transform. The proposed method is found to have salient features namely, less human interaction, significant reduction in computational speed and provides very precise image registration with sub-pixel accuracy when compared with the traditional approaches.