Multi-Scale Neural BRDFs for Heterogeneous Volume and Surface Rendering

While Zhou et al. (2024) introduce a physically-based neural BRDF (PBNBRDF) for surfaces, and Igouchkine et al. (2018) tackle multi-material volume rendering, no current model smoothly integrates both. Here, the idea is to develop a neural representation that respects physical laws (reciprocity, energy conservation) across both surfaces and volumes, interpolating material properties at interfaces. This would enable, for example, accurate rendering of foggy glass, muddy water, or biological tissues—where both scattering and surface reflections are critical. The innovation is a unified, scale-adaptive neural model, possibly leveraging the hierarchical Gaussian structures of Kerbl et al. (2024), that can transition between surface and volume regimes for complex, real-world materials.

References:

1. A Hierarchical 3D Gaussian Representation for Real-Time Rendering of Very Large Datasets. Bernhard Kerbl, Andreas Meuleman, Georgios Kopanas, Michael Wimmer, Alexandre Lanvin, G. Drettakis (2024). ACM Transactions on Graphics.
2. Physically Based Neural Bidirectional Reflectance Distribution Function. Chenliang Zhou, Alejandro Sztrajman, Gilles Rainer, Fangcheng Zhong, Fazilet Gokbudak, Zhilin Guo, Weihao Xia, Rafal Mantiuk, Cengiz Oztireli (2024). arXiv.org.
3. Multi-Material Volume Rendering with a Physically-Based Surface Reflection Model. Oleg Igouchkine, Yubo Zhang, K. Ma (2018). IEEE Transactions on Visualization and Computer Graphics.