Efficient voxelization using projected optimal scanline

In the paper, we propose an efficient algorithm for the surface voxelization of 3D geometrically complex models. Unlike recent techniques relying on triangle-voxel intersection tests, our algorithm exploits the conventional parallel-scanline strategy. Observing that there does not exist an optimal scanline interval in general 3D cases if one wants to use parallel voxelized scanlines to cover the interior of a triangle, we subdivide a triangle into multiple axis-aligned slices and carry out the scanning within each polygonal slice. The theoretical optimal scanline interval can be obtained to maximize the efficiency of the algorithm without missing any voxels on the triangle. Once the collection of scanlines are determined and voxelized, we obtain the surface voxelization. We fine tune the algorithm so that it only involves a few operations of integer additions and comparisons for each voxel generated. Finally, we comprehensively compare our method with the state-of-the-art method in terms of theoretical complexity, runtime performance and the quality of the voxelization on both CPU and GPU of a regular desktop PC, as well as on a mobile device. The results show that our method outperforms the existing method, especially when the resolution of the voxelization is high.