Jaroslav Křivánek

Monte Carlo methods for physically based volume rendering

SIGGRAPH 2018 Course

Jan Novák Disney Research

Iliyan Georgiev Solid Angle

Johannes Hanika Karlsruhe Institute of Technology

Jaroslav Křivánek Charles University, Prague Render Legion | Chaos Group

Wojciech Jarosz Dartmouth College

(hovering mouse over a presenter's name shows his bio sketch)

Various media rendered with Monte Carlo methods for physically based simulation of light transport in volumes. The three images on the left are courtesy of Lee Griggs.

Abstract

We survey methods that utilize Monte Carlo (MC) integration to simulate light transport in scenes with participating media. The goal of this course is to complement a recent EUROGRAPHICS 2018 state-of-the-art report providing a broad overview of most techniques developed to date, including a few methods from neutron transport, with a focus on concepts that are most relevant to CG practitioners. The wide adoption of path-tracing algorithms in high-end realistic rendering has stimulated many diverse research initiatives aimed at efficiently rendering scenes with participating media. More computational power has enabled holistic approaches that tie volumetric effects and surface scattering together and simplify authoring workflows. Methods that were previously assumed to be incompatible have been unified to allow renderers to benefit from each method's respective strengths. Generally, investigations have shifted away from specialized solutions, e.g. for single- or multiple-scattering approximations or analytical methods, towards the more versatile Monte Carlo algorithms that are currently enjoying a widespread success in many production settings. The goal of this course is to provide the audience with a deep, up-to-date understanding of key techniques for free-path sampling, transmittance estimation, and light-path construction in participating media, including those that are presently utilized in production rendering systems. We present a coherent overview of the fundamental building blocks and we contrast the various advanced methods that build on them, providing attendees with guidance for implementing existing solutions and developing new ones.... Extended abstract

Reference

Jan Novák, Iliyan Georgiev, Johannes Hanika, Jaroslav Křivánek, and Wojciech Jarosz. Monte Carlo methods for physically based volume rendering. ACM SIGGRAPH 2018 Courses (SIGGRAPH '18). ACM, New York, NY, USA.
DOI | BibTeX

Presented on Wednesday, 15 August 2018, 9am - 12:15pm in West Building, Room 301-305, Vancouver Convention Centre.

Course Notes

1.	Introduction (Wojciech Jarosz) slides (pdf)
2.	Fundamentals (Jan Novák) slides (keynote) | slides (pdf) | notes pages
3.	Distance sampling (Jan Novák) slides (keynote) | slides (pdf) | notes pages
4.1.	Transmittance estimation (Wojciech Jarosz) slides (pdf)
4.2.	Transmittance using null collisions (Jan Novák) slides (keynote) | slides (pdf) | notes pages
5.	Path construction (Iliyan Georgiev) slides (keynote) | slides (pdf)
6.1.	Advanced methods (Wojciech Jarosz) slides (pdf)
6.2.	Advanced methods: Combining estimators (Jaroslav Křivánek) slides (pptx) | slides (pdf) | notes pages
6.3.	Advanced methods: Zero-variance-based sampling and path guiding (Jaroslav Křivánek) slides (pptx) | slides (pdf) | notes pages
6.4.	Advanced methods: Acceleration structures (Johannes Hanika) slides (pdf) |
7.	Open problems and challenges (Jaroslav Křivánek) slides (pptx) | slides (pdf) | notes pages

Acknowledgements