Multiresolution image registration

Abstract

The paper describes an automatic registration procedure based on a multiresolution analysis of images. The approach is quite general and can be applied to a large variety of images. Furthermore the algorithm is very robust and can effectively cope with a considerable range of transformations, since the registration is obtained iteratively at different multiresolution scales. The procedure is completely automatic, and relies on the grey level information content of the images. We have applied the algorithm to test images of banknotes, aerial stereo pairs, multispectral and SAR images. In all the cases we have obtained excellent results, outperforming the best algorithms available at present and used in industrial applications.