基于形态学商图像的光照归一化算法(英文)  

Illumination Normalization with Morphological Quotient Image

作　　者：何晓光[1] 田捷[1] 毋立芳[2] 张瑶瑶[1] 杨鑫[1] 

机构地区：[1]中国科学院自动化研究所复杂系统与智能科学实验室,北京100080 [2]北京工业大学电子信息与控制工程学院,北京100022

出　　处：《软件学报》2007年第9期2318-2325,共8页Journal of Software

基　　金：Supposed by the Project of National Science Foundation for Distinguished Young Scholars of China under Grant No.60225008(国家杰出青年科学基金);the Key Project of National Natural Science Foundation of China under Grant Nos.603320lO,60575007(国家自然科学基金重点项目);the Project for Young Scientists' Foundation of National Natural Science of China under Grant No.60303022(国家自然科学基金青年科学基金);the Project of Natural Science Foundation of Beijing of China under Grant No.4052026(北京市自然科学基金);the Beijing Municipal Education Commission Foundation of China under Grant No.KM200610005011(北京市教委基金)

摘　　要：复杂光照条件下的人脸识别是一个困难但需迫切解决的问题,为此提出了一种有效的光照归一化算法.该方法根据面部光照特点,基于数学形态学和商图像技术对各种光照条件下的人脸图像进行归一化处理,并且将它发展到动态地估计光照强度,进一步增强消除光照和保留特征的效果.与传统的技术相比,该方法无须训练数据集以及假定光源位置,并且每人只需一幅注册图像.在耶鲁人脸图像库B上的测试表明,该算法以较小的计算代价取得了优良的识别性能.Face recognition under complex illumination conditions is still an open question. To cope with the problem, this paper proposes an effective illumination normalization method. The proposed method employs morphology and quotient image techniques by analyzing the face illumination, and it is upgraded with dynamical lighting estimation technique to strengthen illumination compensation and feature enhancement. Compared with traditional approaches, this method doesn＇t need any training data and any assumption on the light conditions, moreover, the enrollment requires only one image for each subject. The proposed methods are evaluated on Yale Face database B and receive a very comoetitive recognition rate with low computational cost.

关 键 词：人脸识别 光照归一化 形态学 商图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]