Technical Papers
Sampling, Reconstructing, and Filtering Light
Tuesday, 7 August 10:45 AM - 12:15 PM | Los Angeles Convention Center, Room 502AB
Session Chair: Kun Zhou, Zheijang University
Sampling, Reconstructing, and Filtering Light
A Theory of Monte Carlo Visibility Sampling
A theory of Monte Carlo visibility sampling, using statistical and pixel-light Fourier analysis. For a linear light, this paper shows that a variant called uniform jitter sampling is nearly optimal. For planar area lights, uniform jitter is better on circular lights, and stratified is better on square lights.
Ravi Ramamoorthi
University of California, Berkeley
John Anderson
Pixar Animation Studios
Mark Meyer
Pixar Animation Studios
Derek Nowrouzezahrai
Disney Research Zürich and Université de Montréal
Theory, Analysis, and Applications of 2D Global Illumination
This paper derives a theory of 2D global illumination, which accelerates experimentation and teaching, and provides insights for 3D rendering. The paper demonstrates this by developing several novel improvements to the irradiance caching algorithm.
Wojciech Jarosz
Disney Research Zürich and University of California, San Diego
Volker Schönefeld
Limbic Software, RWTH Aachen University, and University of California, San Diego
Leif Kobbelt
RWTH Aachen University
Henrik Wann Jensen
University of California, San Diego
On Filtering the Noise From the Random Parameters in Monte Carlo Rendering
A new way to identify and filter Monte Carlo noise by estimating the functional relationship between sample values and the random parameters used in the rendering system.
Pradeep Sen
University of New Mexico
Soheil Darabi
University of New Mexico
Reconstructing the Indirect Light Field for Global Illumination
A general reconstruction technique that dramatically reduces noise in global illumination renderings by exploiting anisotropy in the indirect light field. This permits efficient reuse of samples between pixels or world-space locations.
Jaakko Lehtinen
NVIDIA Research
Timo Aila
NVIDIA Research
Samuli Laine
NVIDIA Research
Frédo Durand
Massachusetts Institute of Technology