Spatial partitioning of geometry images using locality masks

The advantages of using geometry images as surface representations largely depend on their regular sampling distribution and strictly ordered 2D storage arrangement. Traditional 3D spatial partitioning techniques often compromise these attractive properties when building hierarchical data structures. We present a modification to traditional partitioning methods using locality masks, which maintain the original sampling and storage structure of geometry images. Applications using spatial hierarchies can then take advantage of the sequential memory access and simplified sampling neighbourhoods associated with geometry images without an intermediate sorting phase. The method uses traditional principles for creation, storage and processing of internal hierarchy nodes, but treats the referencing of primitives at leaf nodes differently. Locality masks are presented with future geometry image processing techniques in mind and handle both single and multi-chart geometry images.