Hermes, JanHenrich, NiklasGrosch, ThorstenMueller, StefanReinhard Koch and Andreas Kolb and Christof Rezk-Salama2014-02-012014-02-012010978-3-905673-79-1https://doi.org/10.2312/PE/VMV/VMV10/065-072A fast computation of unbiased global illumination is still an unsolved problem, especially if multiple bounces of light and non-diffuse materials are included. The standard Monte Carlo methods are time-consuming, because many incoherent rays are shot into the scene, which is hard to parallelize. On the other hand, GPUs can make the most of their computing power if the problem can be broken down into many parallel, small tasks. Casting global, parallel ray-bundles into the scene is a way of achieving this parallelism. We exploit modern GPU features to extract all intersection points along each ray within a single rendering pass. Radiance can then be transferred between pairs of all points which allows an arbitrary number of interreflections, especially for compelling multiple glossy reflections. Beside arbitrary BRDFs, our method is independent of the number of light sources and can handle arbitrary shaped light sources in a unified framework for unbiased global illumination. Since many methods exist for fast computation of direct light using soft shadows, we demonstrate how our method can be built on top of any direct light simulation.Categories and Subject Descriptors (according to ACM CCS): I.3.3 [Computer Graphics]: Picture/Image Generation