Most occlusion culling algorithms select a subset of suitable occluder planes or geometries to exclude invisible objects from further visualization processing. Preferably these occluders are large and simple. For complex scenes it is worthwhile to generate virtual occluders to replace complex occluder geometries by simple polygonal structures. In urban scenes, the facade of buildings comprises most of the occlusion potential. In this paper we present an algorithm to extract simplified facades from complex building models, exploiting the fact that, in most cases, buildings have a 2.5D structure.
Rick Germs, Frederik W. Jansen