Terrain rendering based on virtual attention

In this paper we present a novel method for terrain rendering which is based on visual attention. Our method first statically subdivides the terrain into some rectangular patches that is dynamically subdivided into some rectangular tiles at runtime. Finally, the tiles are used to construct mesh based on visual attention. Because we calculate the levels of detail of each tile's edges and the adjacent tiles share a same edge, the resulted mesh does not include any cracks naturally. Dynamically subdivision can balance the number of calling API and the complexity of geometry to improve the efficiency of rendering. Visual attention not only makes the render quality is better, but also stabilize the frame rate and eliminate the visual bound. Our approach achieves terrain rendering at a high stable frame rate.