Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
508032 | Computers & Geosciences | 2010 | 8 Pages |
Remote sensing technologies, such as LIDAR, rapidly evolve and produce large datasets. The computers used to visualize these data have limited resources, which prevent detailed and real-time visualization. An approach to real-time visualization of virtually unlimited LIDAR datasets, at full detail with a hierarchical and out-of-core approach to data management and a modern point-based rendering technique, is presented. It is based on on-demand loading of data subsets into their optimal memory locations and on a two-pass point-based rendering method that utilizes elliptical weighted average filtering to reduce alias. In addition, points are rotated in accordance to their distance and orientation from the viewer. All graphics computations are implemented in programmable shaders on the GPU, so the CPU is free to perform other tasks. The capability of our approach is compared with the Kreylos et al. (2008) method.