A Rapid Clustering Algorithm for Efficient Rendering

Abstract

Hierarchical radiosity using object clusters greatly improves rendering times and reduces memory consumption of radiosity computations. The key feature of the algorithm is using a hierarchy of object clusters to approximate the energy exchange between surfaces. The cluster hierarchy used for this purpose however, must accurately reflect the actual scene geometry to justify this approach. Bad clusters easily lead to rendering artifacts. Inspired by the results of our hierarchical bounding volume optimization for ray tracing, we applied the same scheme to a clustering algorithm for hierarchical radiosity. Using an object-oriented framework, the extension of the implementation was straight forward and seems to be promising. Due to the good performance of our hierarchy regarding ray tracing, the same data structure could successfully be used for two things: visibility checks based on ray casting and energy exchange for the radiosity computation. In this short paper, first results regarding our new clustering scheme will be presented. The properties of the underlying bounding volume optimization give reason for interesting applications enhancing various rendering techniques. Some of these new ideas will be discussed here.