Real-time Global Illumination using Resampling Methods

• Applicant: Professor Dr. Thorsten Grosch
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 560374085

Today, Monte Carlo methods are often used to simulate global illumination, where the integral equation is solved numerically using randomly generated paths. Since ray tracing hardware is available for some years now, the simulation can be displayed at interactive speed, but still shows noisy images. Using the ReSTIR method (Reservoir-based Spatio-temporal Importance Resampling), some first real-time capable global illumination simulations are possible. Here, a complex sampling function is approximated in several stages by simple sampling functions, and by reusing the illumination results several times in reservoirs, the noise can be significantly reduced. In this project, the research in the field of ReSTIR methods should be continued. One focus is on real-time-capable bidirectional Monte Carlo techniques that further improve the image quality by simulating additional light paths. Additionally, real, complex light sources made out of glass and mirrors should be integrated in the simulation, and the illumination of large, extended scenes should be possible without visible loss of quality.

• DFG Programme: Research Grants