Item

Abstract

The paper consists of two main parts: presentation of an efficient global illumination algorithm and description of its extensive experimental validation. In the first part, a hybrid of cluster-based hierarchical and progressive radiosity techniques is proposed, which does not require storing links between interacting surfaces and clusters. The clustering does not rely on input geometry, but is performed on the basis of local position in the scene for a pre-meshed scene model. The locality of the resulting clusters improves the accuracy of form factor calculations, and increases the number of possible high-level energy transfers between clusters within an imposed error bound. Limited refinement of the hierarchy of light interactions is supported without compromising the quality of shading when intermediate images are produced immediately upon user request. In the second part, a multi-stage validation procedure is proposed and results obtained using the presented algorithm are discussed. At first, experimental validation of the algorithm against analytically-derived and measured real-world data is performed to check how calculation speed is traded for lighting simulation accuracy for various clustering and meshing scenarios. Then the algorithm performance and rendering quality is tested by a direct comparison of the virtual and real-world images of a complex environment.