A robust subset-ICP method for point set registration

Abstract

Iterative Closest Point (ICP) is a popular point set registration method often used for rigid registration problems. Because of all points in ICP-based method are processed at each iteration to find their correspondences, the method's performance is bounded by this constraint. This paper introduces an alternative ICP-based method by considering only subset of points whose boundaries are determined by the context of the inputs. These subsets can be used to sufficiently derive spatial mapping of point's correspondences between the source and target set even if points have been missing or modified slightly in the target set. A brief description of this method is followed by a comparative analysis of its performance against two ICP-based methods, followed by some experiments on its subset's sensitivity and robustness against noise. © 2013 Springer International Publishing.