Two algorithms for efficient and realistic rendering
Speaker: Matthias Zwicker, University of Bern, Switzerland
Date: Wednesday, August 22 2012
Time: 4:00PM to 5:00PM
Refreshments: 3:50PM
Location: 32-D507
Host: Fredo Durand, MIT: CSAIL: Computer Graphics Group
Contact: Britton 'Bryt' Bradley, 617-253-6583, bryt@csail.mit.edu
Relevant URL: Abstract:
Monte Carlo integration is widely considered the most practical approach
for realistic rendering in computer graphics. In many practical
scenarios, however, Monte Carlo techniques are hampered by noise
artifacts and slow convergence to visually pleasing results. In this
talk I will present two algorithms that we recently developed in my
research group addressing this issue. The first one is a progressive
photon mapping technique. Our approach renders low noise images in many
situations that are challenging for other techniques. In addition, it
also progressively reduces bias or systematic error, which has been a
downside of previous photon mapping techniques. In the second part of
the talk I will describe an image-based adaptive sampling and
reconstruction technique that attempts to estimate and minimize the
error given a certain sample budget. I will show how this approach,
despite its simplicity, significantly reduces error and improves visual
quality compared to standard techniques.
See other events happening in August 2012
