Improved Billboard Clouds for Extreme Model Simplification

Computer generated scenes are usually represented by polygon models. While graphics hardware capabilities have advanced rapidly in recent years many natural scenes are still much too complex to be rendered in real-time as polygon models. A solution is to simplify such models by reducing the number of polygons in them. However, for many complex models, such as models of trees, a polygon reduction is difficult to achieve. Recently a new image-based method called “Billboard Clouds” has been suggested for extreme model simplification. The idea is to replace a complex model by a set of texture mapped images (billboards) of it. The algorithm can deliver impressive results but has several severe limitations: the set of possible planes is restricted to a discretised plane space and the number of resulting billboards can not be controlled directly. In this paper we present a new algorithm for generating billboard clouds using k-means clustering. We show that our algorithm is fast and offers improved performance and user control.