Figure 1: In this scene, we use a near-UV light for illumination. While non-fluorescent materials only reflect shades of the illumination colour (left), fluorescent surfaces also reemit a portion of the absorbed energy from incident light as a light at additional wavelengths, leading to the colourful appearance of other objects in the scene. To represent fluorescent materials in a renderer, we typically use reradiation matrices (centre), which have a significant memory overhead. Instead, we propose a more efficient representation of these matrices using Gaussian mixture models (right -eight Gaussians). Although compact, this representation also provides compelling results even with such challenging illumination.

Monte Carlo rendering makes heavy use of mixture sampling and multiple importance sampling (MIS). Previous work has shown that control variates can be used to make such mixtures more efficient and more robust. However, the existing approaches failed to yield practical applications, chiefly because their underlying theory is based on the unrealistic assumption that a single mixture is optimized for a single integral. This is in stark contrast with rendering reality, where millions of integrals are computed---one per pixel---and each is infinitely recursive. We adapt and extend the theory introduced by previous work to tackle the challenges of real-world rendering applications. We achieve robust mixture sampling and (approximately) optimal MIS weighting for common applications such as light selection, BSDF sampling, and path guiding.