Fast Parallel Surface and Solid Voxelization on GPUs

摘要

This paper presents data-parallel algorithms for surface and solidrnvoxelization on graphics hardware. First, a novel conservativernsurface voxelization technique, setting all voxels overlapped by arnmesh’s triangles, is introduced, which is up to one order of magnitudernfaster than previous solutions leveraging the standard rasterizationrnpipeline. We then show how the involved new triangle/boxrnoverlap test can be adapted to yield a 6-separating surface voxelization,rnwhich is thinner but still connected and gap-free. Complementingrnthese algorithms, both a triangle-parallel and a tile-basedrntechnique for solid voxelization are subsequently presented. Finally,rnaddressing the high memory consumption of high-resolutionrnvoxel grids, we introduce a novel octree-based sparse solid voxelizationrnapproach, where only close to the solid’s boundary finestlevelrnvoxels are stored, whereas uniform interior and exterior regionsrnare represented by coarser-level voxels. This representationrnis created directly from a mesh without requiring a full intermediaternsolid voxelization, enabling GPU-based voxelizations of unprecedentedrnsize.