Neural Free‐Viewpoint Relighting for Glossy Indirect Illumination

Precomputed Radiance Transfer (PRT) remains an attractive solution for real-time rendering of complex light transport effects such as glossy global illumination. After precomputation, we can relight the scene with new environment maps while changing viewpoint in real-time. However, practical PRT methods are usually limited to low-frequency spherical harmonic lighting. All-frequency techniques using wavelets promising but have so far had little impact. The curse dimensionality and much higher data requirements typically them relighting fixed view or only direct lighting triple product integrals. In this paper, demonstrate a hybrid neural-wavelet high-frequency indirect illumination, including reflection, view. Specifically, seek represent function Haar wavelet basis. For learn small multi-layer perceptron (MLP) applied feature field spatial location index, reflected direction material parameters being other MLP inputs. We optimize/learn (compactly represented by tensor decomposition) from multiple images under different viewing conditions. (512 x 512 at 24 FPS, 800 600 13 FPS) precomputed challenging scenes involving view-dependent reflections even caustics.