Realtime visibility culling for the visualization and transmission of large 3D scenes

In this paper a new visibility culling is proposed to visualize large, dynamic and interactive 3D scenes in realtime. It makes use of OpenGL and is divided into two phases, namely view frustrum culling and occlusion culling. The occlusion culling itself is divided into an occluder selection and an occlusion processing. The presented approach makes use of spatial, image-based and temporal coherences. The algorithm's precisision can be adjusted by five parameters in order to adapt the culling effort to the computation power of the computer system during runtime.