摘要： 分析了Retinex及其与商图像(QuotientImage,QI)的关系,提出了自商图像(self-quotientimage)的概念,以提取一个目标(如人脸)图像的内在特性,同时去除与光照相关的外在因素。SQI(self-quotientimage)方法克服了原始的QI方法的局限性,且能够提取与光照无关的内在特性。对算法适用的条件进行了理论分析,并在SQI计算过程中提出了一个非迭代的各向异性滤波器以减小光晕现象和提高计算速度。实验结果表明该方法可以有效地提高不同光照条件下的人脸识别系统的识别率。

关键词: 光线变化, 自商图像, 人脸识别, 商图像

Abstract: This paper presents an analysis of Retinex and its relations to the quotient image (QI),and proposes a novel concept,called the self-quotient image (SQI),for extraction of intrinsic image property of an object such as the face while removing extrinsic factors corresponding to the illumination.The SQI overcomes limitations of the original QI method and is able to extract the intrinsic property.It performs a theoretical analysis of conditions when the algorithm applies.A non-iterative filtering algorithm is presented for SQI calculation.The experiment results show the effectiveness of the method for face recognition under various lighting conditions.

中图分类号: 

• TP391.41