A Continuous Adjoint Formulation for Radiance Transport

The Monte Carlo approach, as pioneered by Cook et al. [1], Kajiya [4], and Ward et al. [9], samples the directional distributions using \random walks," which in e ect simulate the transport of light by following paths of sample photons. Error is easy to control: as more rays are traced, the variance of the solution is reduced. However, Monte Carlo approaches, while easy to implement and control, are typically slow to converge|and are particularly ill-suited to di use re ectors, which require a large number of rays to sample adequately.