Ray traversal of octree point clouds on the GPU
Aaron Knoll · 2008
We propose an efficient iterative algorithm for ray traversal of an octree on current GPU hardware. From an initial mesh or volume dataset, we construct a balanced octree where each vertex is assigned a leaf node at the lowest possible depth. Interior nodes are averaged from these values, defining levels of detail. We transfer this compact octree to the GPU, and employ our traversal algorithm to render these particles. This yields interactive performance for moderately large (multi-million point) datasets, and renders coarser levels of detail at significantly faster rates. This algorithm is note-worthy in its simple GPU implementation and relative scalability.