Coherent Hierarchical Culling: Hardware Occlusion Queries Made Useful

Abstract We present a simple but powerful algorithm for optimizing the usage of hardware occlusion queries in arbitrary complex scenes. Our method minimizes the number of issued queries and reduces the delays due to the latency of query results. We reuse the results of occlusion queries from the last frame in order to initiate and schedule the queries in the next frame. This is done by processing nodes of a spatial hierarchy in a front‐to‐back order and interleaving occlusion queries with rendering of certain previously visible nodes. The proposed scheduling of the queries makes use of spatial and temporal coherence of visibility. Despite its simplicity, the algorithm achieves good culling efficiency for scenes of various types. The implementation of the algorithm is straightforward and it can be easily integrated in existing real‐time rendering packages based on common hierarchical data structures. Categories and Subject Descriptors (according to ACM CCS): I.3.7 [Computer Graphics]: Three‐Dimensional Graphics and Realism