Real-time Hardware-accelerated Relighting with Approximate Indirect Illumination

Deep framebuffer relighting engines are often used to speed up lighting design in geometricallycomplex procedurally-shaded environments where they provide interactive feedback on changes to the direct illumination. This paper presents an extension to these algorithms by providing real-time feedback for one-bounce indirect illumination. This is achieved by relighting a set of cached gather samples generated from the original geometry using a Monte Carlo gathering approach. To improve performance and decrease storage, the gather samples are clustered such that the resulting data structures are efficient for evaluation on modern GPUs. The hardware-accelerated implementation of our algorithm achieves real-time performance and is scalable to environments with high geometric and material complexity while supporting arbitrary direct lighting models, including local ones, and diffuse and glossy materials.