Adaptive Voxel Subdivision for Ray Tracing

Although regular subdivision has been shown to be efficient at ray tracing scenes where objects are evenly distributed, such algorithms perform poorly when objects are concentrated in a small number of voxels. In this paper, a method is presented where voxels in a regular grid are examined and recursively subdivided depending on object density. This integration of regular and adaptive spatial subdivision methods allows images consisting of large regularly distributed objects and small dense objects to be ray traced efficiently. The parameters controlling the coarseness of the voxel grid, depth of adaptive subdi· vision trees, and maximum number of polygons per voxel are varied and their effects on execution time, subdivision time, and memory use are measured.