|
Speaker |
|
|
Date |
October 25, 2007 |
|
Time |
2:30PM to 3:30PM |
|
Place |
GRAIL (CSE 291) |
A long-term goal of computer graphics has been to produce high-quality
photograph-like or photorealistic images, typically using offline
physically-based rendering algorithms.
Other applications like games, real-time simulations or lighting design
require interactive performance.
Historically, these two worlds have been very separate, with slow high
quality offline rendering, and fast coarse approximations. In this talk, we describe the research we
have done over the last few years to bridge this divide, and obtain high
quality rendering at real-time rates.
The key idea is to develop appropriate compact mathematical
representations of the lighting, reflectance and shadowing or scattering
effects. Many of our mathematical and
computational models, such as those for spherical harmonic environment
lighting, and single scattering for atmospheric effects, are already in
widespread use in production software such as commercial video games.
We discuss three significant projects in this area. First, we describe our early results that
develop the theory for reflection as a spherical convolution of the incident
lighting and reflectance function or BRDF.
These ideas can be used for real-time rendering with natural illumination,
and diffuse or specular materials, as well as for filtering normal mapped
surface detail. Second, we develop a precomputation-based relighting method to include shadows
at all frequencies. This method is based
on a factorization of light transport into material and visibility effects, and
a new framework of wavelet triple product integrals for the product of
lighting, visibility and reflectance.
Third, we address volumetric scattering through a new semi-analytic
single scattering formula. Together,
these ideas have produced some of the most photorealistic interactive images to
date, including complex natural lighting, realistic reflectance, shadows at all frequencies, and physically-based volumetric
scattering.
---------------------------
Ravi Ramamoorthi is currently on the faculty of
Computer Science at Columbia University, where he has been since August 2002
when he received his PhD from Stanford University. He is interested in many aspects of computer
graphics and vision, including mathematical foundations, real-time
photorealistic rendering, image-based and inverse rendering, data-driven
appearance representations and lighting and appearance in computer vision. He received the 2007 ACM SIGGRAPH Significant
New Researcher Award for "his groundbreaking work on mathematical
representations and computational models for the visual appearance of
objects." Earlier this year, he was also named an Office of Naval Research
Young Investigator for his work on "mathematical models of illumination
and reflectance for image understanding and machine vision." Previously, he received a Sloan Research
Fellowship and an NSF Career award in 2005.
A video of his work is available at http://www.cs.columbia.edu/~ravir/RaviR.wmv
---------------------------