Image registration based on Mexican-hat wavelets and pseudo-Zernike moments

Image registration is a key technique in pattern recognition and image processing, and it is widely used in many application areas such as computer vision, remote sensing, image fusion and object tracking. A method for image registration combining Mexican-hat wavelets and pseudo-Zernike moments is proposed. Firstly, feature points are extracted using scale-interaction Mexican-hat wavelets in the reference image and sensed image respectively. Then, pseudo-Zernike moments are used to match them and classical RANSAC used to eliminate the wrong matches. And then, the well match points are used to estimate the best affine transform parameters by least squares minimization. At last, the sensed image is transformed and resampled to accomplish the image registration. The experiments indicate that the proposed algorithm extracts feature points and matches them exactly and eliminates wrong matched points effectively and achieves nice registration results.