Matching point features using mutual information

Abstract

The authors have developed a new mutual information-based registration method for matching unlabeled point features. In contrast to earlier mutual information-based registration methods which estimate the mutual information using image intensity information, the authors' approach uses the point feature location information. A novel aspect of their approach is the spontaneous emergence of correspondence (between the two sets of features) as a natural by-product of information maximization. The authors have applied this algorithm to the problem of geometric alignment of primate autoradiographs. They also present a detailed theoretical comparison between their approach and other approaches that explicitly parameterize feature correspondence.