Image Registration Using Least Square Support Vector Machines

A technique for registration of images with geometric distortions is described. It provides a new insight into the determination of a mapping function by using least square support vector machines (LS-SVM). With this technique, data points are mapped from data space to a high dimensional feature space using a Gaussian kernel in such a way that the mapping function could be evaluated in the feature space. An interesting property of this technique is that it constitutes a practical implementation of the structural risk minimization (SRM) principle that aims at minimizing a bound on the generalization error of a model, rather than minimizing the mean square error over control points. The results of experiments prove that the approach can be devoid of local optima in the optimization process and attain an acceptable generalization result