Neural Histogram-Based Glint Rendering of Surfaces With Spatially Varying Roughness

The complex, glinty appearance of detailed normal-mapped surfaces at different scales requires expensive per-pixel Normal Distribution Function computations. Moreover, large light sources further compound this integration and increase the noise in the Monte Carlo renderer. Specialized rendering techniques that explicitly express the underlying normal distribution have been developed to improve performance for glinty surfaces controlled by a fixed material roughness. We present a new method that supports spatially varying roughness based on a neural histogram that computes per-pixel NDFs with arbitrary positions and sizes. Our representation is both memory and compute efficient. Additionally, we fully integrate direct illumination for all light directions in constant time. Our approach decouples roughness and normal distribution, allowing the live editing of the spatially varying roughness of complex normal-mapped objects. We demonstrate that our approach improves on previous work by achieving smaller footprints while offering GPU-friendly computation and compact representation.

We show a comparison between our method and [GGN18] on four different scenes. Each scene uses a different environment map. Our method has higher quality, uses lower memory and is more performant in all scenes. The teaser scene features a surface with spatially varying roughness which is not supported by [GGN18]. We use ꟻLIP ([ANA*20]) to assess the quality of the results.

1. [ANA*20] Andersson P., Nilsson J., Akenine-Möller T., Oskarsson M., Åström K., Fairchild M. D.: ꟻLIP: A Difference Evaluator for Alternating Images. Proceedings of the ACM on Computer Graphics and Interactive Techniques, 2020, doi:10/gnbnp5
2. [GGN18] Gamboa L. E., Guertin J. P., Nowrouzezahrai D.: Scalable appearance filtering for complex lighting effects. SIGGRAPH Asia 2018. doi:10/ggfg4g.