A Framework of Large-Scale Terrain Visualization Based on GPU

In this paper an efficient solution of large scale, interactive terrain visualization is proposed based on graphic hardware. The method of multi-thread is adopted to resolve the bottle-neck problem of data fetching from the external memory, and the methods of level of details (LODs) based on view-dependent and mesh rendering in graphics programmable units (GPU) are used to ensure visual quality and high rendering frame rate. Our framework exploits restricted quadtree structure to recursively subdivide a square block by the algorithm of top-down refinement, and takes advantage of the character of GPU vertex texture to render surface in shaders which extremely reduce time spending by parallel processing. Experimental results show that the framework is output sensitive, effective and satisfies the requirements of interactive large scale terrain visualization.