Dynamic multiresolution level of detail mesh simplification for real-time rendering of large digital terrain models

There is an increasing demand for high quality interactive 3D visualization of geographical datasets. Because of the inherent geometric complexity and data size, this goal is often unachievable even with high performance graphics workstations, unless the original terrain height field mesh is simplified and approximated in order to reduce the number of geometric primitives that need to be rendered without compromising visual quality. This paper discusses the methodology and implementation aspects of our research work to improve the quality and speed of rendering general purpose desktop PCs by enhancing the geometrical and topological techniques for handing the very large terrain data sets. The proposed algorithm uses a compact and efficient multi-resolution grid representation and employs a variable screen-space threshold to bound the maximum error of the projected image. The appropriate level of detail is computed and generated dynamically in real-time, allowing for smooth changes of resolution across area of the surface. The method is different from the triangle-based LOD algorithms and is optimized for modern, low-end consumer 3D graphics cards and minimizes CPU usage during rendering. The software has been tested on real-world height maps as well as satellite phototextures of size up to 16K/spl middot/16K.The experimental result shows that the proposed algorithm can dynamically generate view-dependent multi-resolution LOD terrain model and real-time rendering can be attained.