PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Learned Color Constancy From Local Correspondences
Tijmen Moerland and Frederic Jurie
In: IEEE International Conference on Multimedia & Expo - 2005(2005).

Abstract

Human color constancy, the ability to correct for color deviation caused by illumination changes, is still unmatched by computer vision performances. The approach taken in this paper learns how illumination changes affect reference objects. It can then explain complex non-linear color transformations with only a few parameters. Global color changes can consequently be estimated from several observations of small reference regions.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Machine Vision
ID Code:1723
Deposited By:Frederic Jurie
Deposited On:28 November 2005