Rendering of Huge Point-Sampled Geometry Based on LOD Control and Out-of-Core Techniques

A new algorithm for interactive rendering of huge point-sampled geometry based on view-dependent level-of-detail （LOD） and out-of-core techniques is presented.At first, point clouds are partitioned into s mall blocks with the same size, and points in each block are reorganized into error-controlled LODs. After that, these LOD data are distributed into main and external memories for balancing systemloads.In the interactive rendering, viewers can control the resolution distribution of points on the whole surfaces by defining current view parameters and using a paging algorithm.Experimental results show that our approach can not only realize real-time rendering for huge datasets, but also is able to improve the usage of main memory for general point clouds.