Image Registration Using Constrained Least Squares

Abstract

This paper presents an image registration algorithm for color images using a constrained least squares(CLS) method. The whole coregistration process formulated around the constraint that there is no zooming but only rotation and translation effects in complex domain between the source and template images. The prior knowledge of the local properties in the images is utilized to create a weight matrix to mitigate the effect of noise and other inhomogeneities. The optimal solution is selected in closed form using least squares to estimate the displacement and rotation parameters which are then used to register the final image.