期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2011
卷号：11
期号：3
页码：139-146
出版社：International Journal of Computer Science and Network Security
摘要：As a color constancy model, this paper proposes a novel method to eliminate illumination effect from images. This method is based on extended gray-world assumption. This assumption states that (1) illumination is not parallel and changes linearly according to positional coordinates, and (2) mean color of objects is gray and color is location unbiased. It is shown that under this assumption illumination can be effectively estimated even if illumination is not parallel then illumination effect is not uniform in the image. Once illumination is estimated, we can estimate the image when white light is illuminated onto the object. Some experimental results are presented, and it is shown that the proposed method effectively eliminate illumination effect even for the case that illumination is not parallel.
关键词：Color constancy; gray-world assumption; object recognition; non-parallel illumination; object color
Loading...
