Higher-Order Radiosity Approximations with a Stochastic Jacobi Iterative Method
Philippe Bekaert | Mateu Sbert | Yves Willems |
Contact: Computer Graphics Research Group
Proceedings of Spring Conference on Computer Graphics 2000, p. 212-221
Budmerice, Slovak Republic (3-6 May)
Abstract
The computation of higher-order polynomial radiosity approximations on a fixed element mesh, results in more smooth images than with a traditional piecewise constant radiosity approximation. Unfortunately, the number of form factors to be stored in a deterministic approach is considerably higher than with a constant approximation and the computation itself of the form factors is more difficult.
In this paper, we present a new stochastic approach for computing higher-order radiosity approximations. The new approach is based on stochastic Jacobi iterations previously used in stochastic ray radiosity and the well-distributed ray set radiosity algorithm. Like in these previous algorithms, explicit computation and storage of form factors is completely avoided. The storage and computation cost of the new algorithm are analysed and several variance reduction techniques are described. In spite of the fact that the new algorithm does not avoid radiosity kernel discretisation, its results are at least as good as with a continuous random walk approach.
Keywords: Radiosity, Monte Carlo method, Stochastic Jacobi iterative method