Fast and Accurate Closest Point Search on Triangulated Surfaces and its Application to Head Motion Estimation

The Iterative Closest Point (ICP) algorithm is widely used to register two roughly aligned surfaces. Its most expensive step is the search for the closest point. Many efficient variants have been proposed to speed up this step, however they do not guarantee that the closest point on a triangulated surface is found. Instead they might result in an approximation of the closest point. In this paper we present a method for the closest point search that is fast and exact. The method was implemented and used to evaluate the accuracy of head motion estimation using dense 3D data obtained from a stereo camera system. The results show that the accuracy of the estimated head motion is better than 1 mm for translational movements and better than 1 degree for rotations. Key-Words: Iterative Closest Point, 3D Registration, Stereo, Head Motion Estimation