Automatic correspondence of range images

Correspondence is a fundamental part of three dimen-sional (3D) model building. It is the process of using sur-face features to determine correspondences between range images. Ideally, we desire a fully automatic system to cre-ate 3D models from range data. However, the automatic correspondence part of the model building procedure has proven very difficult. In this paper we present a nomen-clature to categorise correspondence methods. A variety of methods are reviewed and classified, including classic ex-amples such as iterative closest point matching and more recent techniques such as bitangent curve matching. A new automatic correspondence algorithm, which combines ex-isting techniques with the novel approach of searching the correspondence probability matrix space, is presented. Fi-nally, results for the algorithm are shown for 2D range pro-files.