A multi-view extension of the ICP algorithm

Although the Iterative Closest Point (ICP) algorithm has been an extremely popular method for 3D points or surface registration, it can only be applied to two point sets at a time. By only registering two scans at a time, ICP fails to exploit the redundant information available in multiple scans that have overlapping regions. In this paper, we present a multi-view extension of the ICP algorithm by a method that simultaneously averages the redundant information available in the scans with overlapping regions. Variants of this method that carry out such simultaneous registration in a causal manner and that utilise the transitive property of point correspondences are also presented. The improved accuracy of this motion averaged approach in comparison with ICP and some multi-view methods is established through multiple tests. We also present results of our method applied to some well-known real datasets.