An easy viewer for out-of-core visualization of huge point-sampled models

We propose a viewer for huge point-sampled models by combining out-of-core technologies with view-dependent level-of-detail (LOD) control. This viewer is designed on the basis of a multiresolution data structure we have developed for gaze-guided visualization and transmission of 3D point sets. In order to reduce memory loads for huge point sets on general PC platforms, we introduce a partition-based out-of-core strategy to balance usage of main and external memories. At first, the data surface is partitioned into small blocks and points in each block are reorganized into error-controlled LODs by hierarchical clustering and LOD organization. In the interactive rendering process, a data block scheduling algorithm is used to realize the view-dependent paging. Experimental results show that the viewer can perform interactive visualization of huge point models on commodity graphics platforms with ease.