Surfaces, curves, and points as constraints in nonrigid registration

This paper describes a new method for performing nonrigid registration of brain images where gray level matching is augmented with structural information. Segmented surfaces, curves and points are used as constraints to help increase the accuracy of the registration. Control points are placed in the image only where needed such as at places of high curvature and on segmented structures. The resulting algorithm uses the available information effectively in order to produce an accurate registration.