Abstract:Volume rendering is a costly but widely used technique in scientific visualization. Various methods were proposed to accelerate volume rendering, most of which make use of empty brick skipping to avoid sampling the emtpy space of volumetric data, whereas the empty voxels in the valid bricks are still treated as valid. We introduce Depth-Box VDB, a simple but effective method based on GPU-based voxel database (GVDB) to accelerate volume rendering by further reducing the number of samplings. With the help of depth maps attached to the valid bricks and the displacement mapping technique, our method can skip part of the empty voxels in the valid bricks and has achieved significant improvement in the performance of volume rendering. According to the performance tests on different datasets, the rendering efficiency increases by up to 121.74%, while the use of memory increases by around 18.67%. Comparing with different shading models, we can conclude that our method tends to speed up more as the rendering requires more shading computing.

Depth-Box VDB: Accelerate Sparse Volume Rendering with Depth Maps Through Voxel Database