Fractal Compression Approach for Efficient Interactive Terrain Rendering on the GPU

This paper describes an efficient technique for the rendering of large terrain surfaces. The technique is based on a simple rings structure: a sequence of concentric rings at different resolutions and centeredon the viewer's position. Each ring is represented by a set of patches at identical resolutions. Rings near the viewer have a finer resolution than the rings further from the viewer. At runtime, the patches within the rings change resolution based on the viewer's position. The GPU decodes in real time height maps encoded by a fractal compressor from which sample the height component of the terrain. Since adjacent patches of different rings can disagree on the resolution of common edge GPU stitches the meshes in order to avoid any cracks or degenerate triangles. The renderedmeshes ensure the absence of cracks that may cause the appearance of visual artifacts. In addition, a tile manager support is evaluated in order to maintain terrain datasets on disk storage avoiding a costly load of the entire datasets into the memory.