Optimizing realistic rendering with many-light methods

  • Authors:
  • Jaroslav Krivanek;Milos Hasan;Adam Arbree;Carsten Dachsbacher;Alexander Keller;Bruce Walter

  • Affiliations:
  • Charles University in Prague;University of California, Berkeley;Autodesk, Inc.;Karlsruher Institut für Technologie;NVIDIA Advanced Rendering Center;Cornell University

  • Venue:
  • ACM SIGGRAPH 2012 Courses
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

With recent improvements in hardware performance, there has been an increased demand in various industries, including game development, film production, and architectural visualization, for realistic image rendering with global illumination (GI). But current algorithms can not fulfill the strict speed and quality requirements of modern applications. Many-light rendering solves this problem. By reducing light-transport simulation to rendering the scene with many light sources, the many-light formulation offers a unified view of the global-illumination scene. Unlike other GI algorithms, the quality-speed trade-off in the many-light methods produces artifact-free images in a fraction of a second while converging to the full GI solution over time. This course presents a coherent summary of the state of the art in many-light rendering. It covers the basic many-light formulation and recent work on its use for computing global illumination in real time, on improving scalability with a large number of lights, on using the many-light method as a basis for a full GI solution, and on rendering participating media. The course focuses on the clarity of the underlying mathematical concepts as well as on the practical aspects of the individual algorithms. One segment of the course is devoted to the practical considerations of using many-light methods in the Autodesk Cloud Rendering service.