Point cloud partitioning approach for ICP improvement

In 3D reconstruction applications, an important issue is the matching of point clouds corresponding to different perspectives of a given object in a scene. Traditionally, this problem is solved by the use of the Iterative Closest point (ICP) algorithm. In view of improving the efficiency of this technique, in this paper we propose a preprocessing step which works on the raw point cloud. Additionally, we propose a metrics to evaluate the outcome of the ICP algorithm. Our experiments have been carried out on artificial as well as on real depth maps acquired from a time-of-flight sensor, and revealed that our cloud partitioning approach makes the ICP algorithm to run 25 times faster, at least.