The open-source point cloud library (PCL) and the tools available for point cloud registration is presented. Pairwise registration is usually carried out by means of one of the several variants of the ICP algorithm. Due to the nonconvexity of the optimization, ICP-based approaches require initialization with a rough initial transformation to increase the chance of ending up with a successful alignment. Good initialization also speeds up their convergence. Two major classes of registration algorithms can be distinguished, feature-based registration algorithms (path 1) for computing initial alignments, and iterative registration algorithms (path 2) following the principle of the ICP algorithm to iteratively register point clouds. For the feature-based registration, geometric feature descriptors are computed and matched in some high-dimensional space. The more descriptive, unique, and persistent these descriptors are, the higher is the chance that all found matches are pairs of points that truly correspond to one another. In contrast to the feature-based registration, iterative registration algorithms do not match salient feature descriptors to find correspondences between source and target point clouds, but instead search for closest points (matching step) and align the found point pairs. To speed up registration, another common extension to the original ICP algorithm is to register only subsets of the input point clouds sampled in an initial selection step.