1. Introduction
Rigid registration of point clouds is an important task in 3D shape processing. Given two scans of the same (rigid) object, the goal is to find the transformation that aligns one scan to the other. This has many applications in autonomous driving, 3D reconstruction, medical imaging, etc. In real life, scans are usually noisy and partial, so finding the correct transformation is hard. We consider here the case of rigid transformations with 6 degrees of freedom.