Bayesian Color Constancy for Outdoor Object Recognition - Robotics Institute Carnegie Mellon University

Bayesian Color Constancy for Outdoor Object Recognition

Yanghai Tsin, Robert Collins, Visvanathan Ramesh, and Takeo Kanade
Conference Paper, Proceedings of (CVPR) Computer Vision and Pattern Recognition, pp. 1132 - 1139, December, 2001

Abstract

Outdoor scene classification is challenging due to irregular geometry, uncontrolled illumination, and noisy reflectance distributions. This paper discusses a Bayesian approach to classifying a color image of an outdoor scene. A likelihood model factors in the physics of the image formation process, the sensor noise distribution, and prior distributions over geometry, material types, and illuminant spectrum parameters. These prior distributions are learned through a training process that uses color observations of planar scene patches over time. An iterative linear algorithm estimates the maximum likelihood reflectance, spectrum, geometry, and object class labels for a new image. Experiments on images taken by outdoor surveillance cameras classify known material types and shadow regions correctly, and flag as outliers material types that were not seen previously.

BibTeX

@conference{Tsin-2001-8358,
author = {Yanghai Tsin and Robert Collins and Visvanathan Ramesh and Takeo Kanade},
title = {Bayesian Color Constancy for Outdoor Object Recognition},
booktitle = {Proceedings of (CVPR) Computer Vision and Pattern Recognition},
year = {2001},
month = {December},
pages = {1132 - 1139},
keywords = {Bayes rule, color constancy, material classification, object recognition, learning, image formation model, inference},
}