Layered point clouds

Abstract

We present a simple point-based multiresolution structure for interactive visualization of very large point sampled models on consumer graphics platforms. The structure is based on a hierarchy of precomputed object-space point clouds. At rendering time, the clouds are combined coarse-to-fine with a top-down structure traversal to locally adapt sample densities according to the projected size in the image. Since each cloud is made of a few thousands of samples, the multiresolution extraction cost is amortized over many graphics primitives, and host-to-graphics communication effectively exploits on-board caching and object based rendering APIs. The progressive block based refinement nature of the rendering traversal is well suited to hiding out-of-core data access latency, and lends itself well to incorporate backface, view frustum, and occlusion culling, as well as compression and viewdependent progressive transmission. The resulting system allows rendering of complex models at high frame rates (over 60M splat/second), supports network streaming, and is fundamentally simple to implement.