TITLE: The Canonical Correlations of Natural Color Images and their use for inverse problems (Parts I and II)
SPEAKER: Yacov Hel-Or, HP Labs
DATES:
Part I: 2:00-3:00 P.M., Thursday June 12, 2003
Part II: 11:00-12:00 P.M., Friday June 27, 2003
LOCATION:
Part I in Sigma, 1L (PA)
Part II in Pi, 1L (PA)
HOST: Vinay Deolalikar
ABSTRACT:
This is an informal presentation with the aim of receiving feedbacks and
criticism about an ongoing project which deals with finding a useful
reprentation for the prior of natural color images.
I'll split my presentation into two sessions:
The first session will give background about the Canonical Correlation
Analysis and its use in estimating informative marginal distributions.
In the next session I will show how the canonical correlation representation
can assist in solving inverse problems in color images and specifically its
application to the demosaicing problem.
|
printable version
