Simple and Robust Iterative Importance Sampling of Virtual Point Lights

Abstract

We present a simple and practical algorithm for importance sampling virtual point lights (VPLs) [Kel97], suitable for multi-pass rendering. During VPL distribution, a Russian roulette decision accepts each VPL proportionally to its estimated contribution to the final image. As a result, more VPLs are concentrated in areas that illuminate the visible parts of the scene, at the cost of a negligible performance overhead in the preprocessing phase. As VPLs are sampled independently and proportionally to their camera importance, the algorithm is trivial to parallelize and remains efficient for low sampling rates. We show that this sampling scheme is well suited to both well illuminated scenes as well as for difficult visibility conditions. Moreover, in contrast to bidirectional and Metropolis VPL sampling [SIMP06, SIP07], the algorithm is fast and very simple to implement, and uses a single Monte Carlo sampler, making it easier to maintain good stratification.We present a simple and practical algorithm for importance sampling virtual point lights (VPLs) [Kel97], suitable for multi-pass rendering. During VPL distribution, a Russian roulette decision accepts each VPL proportionally to its estimated contribution to the final image. As a result, more VPLs are concentrated in areas that illuminate the visible parts of the scene, at the cost of a negligible performance overhead in the preprocessing phase. As VPLs are sampled independently and proportionally to their camera importance, the algorithm is trivial to parallelize and remains efficient for low sampling rates. We show that this sampling scheme is well suited to both well illuminated scenes as well as for difficult visibility conditions. Moreover, in contrast to bidirectional and Metropolis VPL sampling [SIMP06, SIP07], the algorithm is fast and very simple to implement, and uses a single Monte Carlo sampler, making it easier to maintain good stratification.