Light probes are an elementary technique for credible visualizations in computer graphics. They can be used to precalculate incident light at various points in the scene, which enables realistic rendering of virtual objects. The VWDS department has already published various papers on the creation of light probes, where reference must be made to special requirements in mixed and mediated reality applications. Other important questions include how many light probes are actually required and where they need to be placed in the scene in order to enable credible lighting in MR applications.

Our current study explores this question and investigates how the positioning of light probes affects visual realism in virtual applications. We present a novel, geometry-based approach for the placement of light probes that takes into account the spatial structure of the 3D scene. A comparative analysis with various previous approaches shows that our method enables a more realistic visualization with fewer or comparable numbers of light probes. The visual advantages over previous approaches were validated with both image metrics and a user study. Virtual environments were used for the comparative data, but the approach can also be applied to real environments. Our method also requires only a short calculation time and is therefore suitable for real-time use. This is particularly relevant for MR applications, where long precalculations are often not possible.

The approach was presented at Cyberworlds 2024 and the publication can be found at the following link: https://ieeexplore.ieee.org/document/10917350