A Clustering Algorithm for Radiance Calculation In General Environments

Abstract

This paper introduces an efficient hierarchical algorithm capable of simulating light transfer for complex scenes containing non-diffuse surfaces. The algorithm stems from a new formulation of hierarchical energy exchanges between object clusters, based on the explicit representation of directional radiometric distributions. This approach permits the simplified evaluation of energy transfers and error bounds between clusters. Representation and storage issues are central to this type of algorithm: we discuss the different choices for representing directional distributions, and the choice between explicit storage or immediate propagation of directional information in the hierarchy. The framework presented is well suited to a multi-resolution representation, which may in turn significantly alleviate the storage problems. Results from an implementation are presented, indicating the feasibility of the approach and its capacity to treat complex scenes.

A unified hierarchical algorithm for global illumination with scattering volumes and object clusters

“…Finally, a number of cluster properties are in fact functions of direction in space, and could be stored as such. The use of such explicitly directional clusters allows constant-time linking operations, but requires important storage capabilities [32].…”

“…Several data structures have been developed for the simulation of nondiffuse surfaces [30] and could be used to store radiance distributions and scattering functions for each cluster. The benefit of this approach would be to maintain the constanttime linking operation, but the tradeoff with the important cost of the directional structure must be carefully investigated [32].…”

Radiance caching for efficient global illumination computation

“…Spherical Harmonics [18], [20], [31], [32], [33], [34], [35], [36], [37] remove the aliasing problem and are efficient for representing low-frequency functions. However, representation of sharp functions requires many coefficients and ringing might appear.…”

Wavelets in computer graphics

“…In order to remove the quadratic dependence on the number of input surfaces the hierarchy of interactions must be extended to scales coarser than the initial set of input surfaces. Algorithms which perform such clustering have recently appeared [56], [55], [7], [54]. The main difficulty with clustering in the context of radiosity is due to visibility.…”