Compact Precomputed Voxelized Shadows

Figure 1: An example of using our algorithm to evaluate precomputed shadows from the sun when viewing the scene at varying scales. Our compact data structure occupies 100MB of graphics memory and is equivalent to a 256k×256k (i.e. 2621442) shadow map. With a filter size of 9×9 taps, shadow evaluation is done in < 1ms at 1080p resolution. Producing high-quality shadows in large environments is an im-portant and challenging problem for real-time applications such as games. We propose a novel data structure for precomputed shadows, which enables high-quality filtered shadows to be reconstructed for any point in the scene. We convert a high-resolution shadow map to a sparse voxel octree, where each node encodes light visibility for the corresponding voxel, and compress this tree by merging common subtrees. The resulting data structure can be many orders of magnitude smaller than the corresponding shadow map. We also show that it can be efficiently evaluated in real time with large filter kernels