Choosing a model of the spatial transformation of images using the criterion of minimum description length

Abstract

This paper discusses the problem of determining the best spatial transformation over the set of reference points found in the process of comparing two images. It is shown that it is allowable to use the rms-deviation criterion only when choosing among transformations with an identical number of parameters. Otherwise, the preference is given to the most complex transformation, which, however, does not provide the best alignment accuracy of the images. A criterion is proposed that is based on the principle of minimum description length for choosing among various classes of transformations (motion, similarity, affine, etc.). It is experimentally established that choosing a spatial transformation on the basis of this criterion makes it possible to increase the alignment accuracy, especially for a small number of reference points or when the positions of the points are given with large errors, as is characteristic of problems of structural alignment of images that are subject to variations.