VIsualization for HPC data - large terrain model

Abstract

Interactive navigation of large terrain data with high-resolution aerial photographic image has been a challenging problem. We intend to conduct visual simulation on large image and geometric data, and create a realistic real-time navigation. We propose the level-of-detail (LOD) modeling with quad-tree (Pajarola, 1998) to speed up the large terrain mesh rendering and using tiled texture (Blythe et al., 1999) to display the detail of the aerial photographic image in realtime rendering. On geometric data, we use LOD and dike structure to provide blocks of meshes with different levels of detail, and remove the possible artifacts and cracks. In respect to the image data, the high-resolution aerial photographic images are divided into tiles, and every tile is pre-processed into multiresolution image data similar to the geometric LOD described above.