Real-Time Illumination of Complex Lights Using Divided Sampling

Existing methods for real-time illumination of complex lights, either require long time pre-computation, or only focus on some special types of illumination.Because computing different kinds of illumination requires different sampling strategies, this paper introduces a novel efficient framework for rendering illumination of complex light sources, in which the pre-computation is not necessary.We divide the rendering equation into three parts: high-frequency term, low-frequency term and occlusion term.Each term is computed by a proper sampling strategy.High-frequency term is solved by importance sampling the BRDF, while low-frequency term is computed by importance sampling the light sources.Occlusion term is computed with depth information in screen-space, and the required number of samples is greatly reduced by interleaved sampling.Our framework is easy to implement on GPU and can solve many real-time rendering problems.The results show that our technique can handle complete light effects with higher quality than previous works.