Point matching is the task of finding correspondences between two arbitrary sets of points. Usually one point set is designated the model set, and the other the data set. Associated with any given set of correspondences is a geometric transformation, which represents the position, or pose, of the model set within the data set.
Our work in point matching has focused on local search as a means of moving through correspondence space. To aid in this process, we make use of a key feature algorithm which can often greatly reduce the time required to find a correct match.
Details of our technique can be found in Jason Denton's dissertation, and Two Dimensional Projective Point Matching, presented at the 5th SSIAI conference. Full citation information in BibTex is available.