Datastructures for capturing dynamic scenes with a time-of-flight camera

Abstract

To capture 3D dynamic scenes, a suitable 3D data structure is needed that can represent the dynamic scene content. In this contribution we analyse and evaluate different data structures for capturing time-varying depth and color data of a dynamic scene obtained with Time-of-Flight and color cameras. The comparison of depth-panoramas, layered depth images and volumetric structures shows that a volumetric octree is best suited to fuse time-varying 3D scene data. We exploit the octree data fusion capabilities for different application scenarios, like 3D environment building, volumetric object reconstruction, and geometric interaction. © 2009 Springer Berlin Heidelberg.