Optimized occlusion culling using five-dimensional subdivision

We describe an algorithm for occlusion culling by visibility preprocessing. The algorithm intelligently partitions the five-dimensional viewing parameter space using a 5D tree (a k-D tree with k = 5), such that each leaf of the tree contains a reference to a subset of the n scene objects. The visibility of each of the objects is determined using a ray-casting procedure. We show how to minimize the number of rays cast in order to achieve a reliable estimate of the visibility relations, and how to efficiently code the resulting 5D tree.