FoldedGI: A highly parallel algorithm for interference detection by folding a geometry image into a 1D buffer

Detecting interference between two or more 3D models plays a critical role in computer animation and virtual reality. Typical approaches are bounding volume hierarchies and spatial partitioning trees, which are mainly carried out on CPUs. The conventional approaches, due to their hierarchical structure, cannot be fully parallelized. In this paper, we propose a highly parallel method, based on geometry images, for detecting interference in real time. Our method is inspired by two important observations - one is that interference between two 3D models can be converted into finding common colors contained in the couple of resulting geometry images, and the other is that the RGB space can be mapped onto a 1D buffer. Our algorithm, called FoldedGI, is parameter free, memory efficient and outperforms the state-of-the-art in terms of speed. We demonstrate its efficacy in dynamic interference detection, penetration depth computation and boolean operations between 3D objects.