Real world scenes often show complex volumetric effects due to scattering of light by particulate media. These effects include some of the most magnificent visual experiences known to man, such as the colors of sunrise and sunset, the brilliant underwater caustics, or even the gloomy weather effects due to fog, mist, haze, rain, snow, and the hazardous effects due to smoke and dust. Realistic visual reproductions of these effects has broad impact in various application domains including digital entertainment, scientific education, simulation and training (flying, diving and medical diagnosis), and safety education (underwater explorations and search efforts in smoky and dusty conditions). While there has been a substantial body of work in computer graphics on rendering the effects of volumetric scattering, those methods are usually too slow to enable widespread adoption, taking hours to days to simulate a single scene. This makes it infeasible to include accurate scattering effects in applications like video games, interactive relighting, virtual environments or scientific education.

The key insight in this research is that while the explicit numerical simulation of scattering is difficult, its qualitative properties are quite simple, usually just involving a blurring or spreading out of light. The investigators first identify simple configurations (or geometries) of media and sources which frequently occur in scenes. They next develop accurate analytic and data-driven models that describe multiple scattering of light in these configurations based on the theory of radiative transfer. These models are then leveraged in conjunction with Monte Carlo, hardware and precomputation methods, to render volumetric effects at interactive rates in a wide range of dynamic scenes and media. Since analytic or algorithmic techniques can only be as accurate as the underlying representations of scattering properties, a critical part of the research involves designing novel experimental setups to measure a wide array of scattering media for the first time. Finally, this effort includes a significant educational component and promotes collaborative spirit, with new joint courses, exchange student visits, visiting lectures, enriching the experiences of both students and researchers.

Agency
National Science Foundation (NSF)
Institute
Division of Computer and Communication Foundations (CCF)
Type
Standard Grant (Standard)
Application #
0541259
Program Officer
Lawrence Rosenblum
Project Start
Project End
Budget Start
2006-02-15
Budget End
2010-01-31
Support Year
Fiscal Year
2005
Total Cost
$225,000
Indirect Cost
Name
Columbia University
Department
Type
DUNS #
City
New York
State
NY
Country
United States
Zip Code
10027