Neural caching improves real time specular lighting quality and speed
Neural Caching of Prefiltered Radiance for Specular Lighting
Graphics
Summary
Creating realistic shiny surfaces in computer graphics is hard because light can reflect in many directions. This paper presents a method where a neural network learns the lighting reflections during rendering, focusing specifically on shiny (specular) surfaces. The network predicts how light reflects depending on surface roughness and direction, which helps the image look better faster compared to previous methods. The method works in real time and improves the quality of shiny reflections in scenes like a bunny and a hall model.
What this means in practice
- •For real time graphics developers: Improve the speed and quality of shiny surface reflections during game or simulation rendering processes.
- •For visual effects studios: Enhance interactive previews of specular lighting in complex scenes for faster iteration and higher fidelity in film rendering pipelines.
Authors
Dmitrii Klepikov, Vladimir Frolov
Abstract
Neural Radiance Caching (NRC) provides an online neural representation of scene illumination for real-time path tracing. This paper presents an NRC variant tailored to specular lighting through a reflection-direction parameterization and a roughness-dependent radiance target. Our network predicts prefiltered incoming radiance at individual surface points, which is combined with a precomputed BRDF integration map using the split-sum approximation to estimate outgoing radiance. On Bunny and Specular Sponza, the reported comparisons indicate faster convergence to the accumulated path-tracing reference and improved specular-lighting quality relative to the evaluated NRC baselines. Our online-trained method maintains real-time performance while adapting its cache during rendering.