Multiresolution image registration

The work describes how the application of multiresolution techniques improves the performance of an image registration algorithm. The algorithm determines the parameter values of a global geometric transformation that corrects gross misregistration between two images. It is assumed that the only distortion present is geometric and that there is no systematic radiometric distortion, i.e. any differences in intensity are due to noise, and not to overall brightness or contrast differences. The method is based upon that of Dana and Anandan (1993), who use a similar technique to register binary edge images for multi-modal registration