Improved Path Guiding for Subsurface Light Scattering

• Thesis title in Czech: Vylepšené navádění cest pro podpovrchový rozptyl světla
• Thesis title in English: Improved Path Guiding for Subsurface Light Scattering
• Key words: fotorealistická syntéza obrazku|sledování paprsku|navádění Monte Carlo integrace
• English key words: photorealistic computer graphics|path tracing|guiding of Monte Carlo integration
• Academic year of topic announcement: 2021/2022
• Thesis type: dissertation
• Thesis language: angličtina
• Department: Department of Software and Computer Science Education (32-KSVI)
• Supervisor: doc. Alexander Wilkie, Dr.
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 05.04.2022
• Date of assignment: 05.04.2022
• Confirmed by Study dept. on: 05.04.2022

Guidelines

The simulation of materials that exhibit subsurface scattering (SSS) properties has become ubiquitous in production renderers due to the increased visual realism they offer [2, 3, 1]. Brute force Monte Carlo approaches average light transport along traced paths with many vertices inside the volume, which is considered the gold standard in the visual effects industry due to its ability to deal with almost arbitrary material configurations: however, this approach is so computationally intensive that its usage has to be restricted to cases where it is absolutely necessary. This has led to significant research into simplifying the underlying mathematical framework, or to provide analytical solutions.
Importance sampling of the SSS walk based on zero-variance theory is one such approach that originates from reactor shield simulations in neutron transport. The idea is that rays entering the medium should not in-scatter deeply, and the ray is guided back to exit the material from the incident half-space. This assumes the importance is spread uniformly on the surface above the medium, and that the surface is a half-slab. This approach, known as Dwivedi guiding [5], can significantly improve the convergence of images with subsurface scattering materials when the simplifying assumptions hold. Whilst further improvements to this technique exist [6, 4], there is no consideration of the global radiance field as the source of importance. Additionally, Dwivedi guiding has not been applied to anisotropic scattering due to the more complicated nature of the directional sampling[4].
Variance reduction in Monte Carlo ray tracing has been achieved through various importance sampling techniques, or their combination via multiple importance sampling. However, many of these rely on local properties of the interaction they are in such as the Dwivedi guiding described above. Radiance field learning has become an essential tool in recent years, its objective being in reducing variance by learning about the global radiance field in the whole scene. Thus, light transport algorithms can reduce variance by guiding rays to parts of the scene where the contributions are more significant. This classification of techniques is known as path guiding and has garnered significant attention in both the research and production community for their effectiveness.
In our research, we will explore the application of path guiding to increase efficiency of sampling full subsurface scattering paths. The learnt radiance field can be used to lift the assumption that the importance is spread uniformly on the surface, hence guide rays in directions where contributions are more significant. Furthermore we would like to experiment with guiding in conjunction anisotropic phase functions, an unexplored area of research as of yet.

References

[1] Per Christensen, Julian Fong, Jonathan Shade, Wayne Wooten, Brenden Schubert, Andrew Kensler, Stephen Friedman, Charlie Kilpatrick, Cliff Ramshaw, Marc Bannister, Brenton Rayner, Jonathan Brouillat, and Max Liani. Renderman: An advanced path-tracing architecture for movie rendering. ACM Trans. Graph., 37(3), August 2018.
[2] Luca Fascione, Johannes Hanika, Mark Leone, Marc Droske, Jorge Schwarzhaupt, Tomáš Davidovič, Andrea Weidlich, and Johannes Meng. Manuka: A batch-shading architecture for spectral path tracing in movie production. ACM Trans. Graph., 37(3), August 2018.
[3] Iliyan Georgiev, Thiago Ize, Mike Farnsworth, Ramón Montoya-Vozmediano, Alan King, Brecht Van Lommel, Angel Jimenez, Oscar Anson, Shinji Ogaki, Eric Johnston, Adrien Herubel, Declan Russell, Frédéric Servant, and Marcos Fajardo. Arnold: A brute-force production path tracer. ACM Trans. Graph., 37(3), August 2018.
[4] Alexander Keller, Pascal Grittmann, Jiří Vorba, Iliyan Georgiev, Martin Šik, Eugene d’Eon, Pascal Gautron, Petr Vévoda, and Ivo Kondapaneni. Advances in monte carlo rendering: The legacy of Jaroslav Křivánek. In ACM SIGGRAPH 2020 Courses, SIGGRAPH ’20, New York, NY, USA, 2020. Association for Computing Machinery.
[5] Jaroslav Křivánek and Eugene d’Eon. A zero-variance-based sampling scheme for monte carlo subsurface scattering. In ACM SIGGRAPH 2014 Talks, SIGGRAPH ’14, pages 66:1–66:1, New York, NY, USA, 2014. ACM.
[6] Johannes Meng, Johannes Hanika, and Carsten Dachsbacher. Improving the Dwivedi sampling scheme. Computer Graphics Forum (Proceedings of Eurographics Symposium on Rendering), 35(4):37–44, June 2016.