Hierarchical clustering of large volumetric datasets

Abstract

In this paper we propose a multiresolution hierarchical data structure called the Ordered Cluster Binary Tree (OCBT). The OCBT is a binary tree structure that extends a Cluster Binary Tree with spatial splitting similar to that of a k-D Tree. We also show how this tree can be improved to extract data efficiently at different sub volumes and levels of detail at run time. We also incorporate a bounding sphere hierarchy to enable early search termination. This clustering algorithm can be made out-of-core and thus enables datasets of several giga bytes in size.