Automatic and robust image registration using feature points extraction and Zernike moments invariants

In this paper a new image registration algorithm is proposed. Rotation, translation, and scaling (RTS) transformations are considered in the registration process. The proposed algorithm consists of three main steps: extraction of some feature points using a robust feature points extractor based on scale-interaction of Mexican-hat wavelets, obtaining the correspondence between the features points of the reference and distorted images based on using Zernike moments of neighbourhoods centered on feature points, and estimating the transformation parameters mapping the distorted image to the reference one. Experimental results illustrate the registration accuracy of the proposed technique and its robustness against several common image-processing operations