自商图像优化加权LBP直方图的光照变化人脸识别  

Illumination Variation Face Recognition Based on Self Quotient Image and Weighted LBP Histogram

作　　者：王刘涛[1] 李肖立 

机构地区：[1]湖南大学信息科学与工程学院,长沙410082

出　　处：《科学技术与工程》2014年第28期98-103,117,共7页Science Technology and Engineering

基　　金：国家自然科学基金重点项目(61133005);国家自然科学基金资助项目(61170263)资助

摘　　要：针对光照变化环境下人脸识别难度大的问题,提出了一种基于自商图像(SQI)和加权局部二值模式(LBP)的光照变化人脸识别方法。首先,利用SQI方法消除光照变化;然后,将两个加权的多区域LBP分别应用于各个图像以提取光照不变特征;最后,将自商图像方法和加权LBP算法提取的直方图融合形成最终的特征向量并用于人脸识别。在扩展YaleB和CMU-PIE人脸数据库上的实验结果表明,相比其他几种人脸识别方法,取得了更好的识别效果。To solve the problem of face recognition in illumination invariant situation, a novel illumination invariant face recognition approach based on Self Quotient Image and weighted Local Binary Pattern is proposed. Firstly, SQI is used for eliminating illumination variation. Then two weighted multi-region LBP with different regions count were applied to face image separately and used for illumination invariant feature extraction. Finally, the face recognition system is presented by combing the Self Quotient Image and weighted LBP operator. The proposed approach is compared with some other methods on extended YaleB and CMU-PIE database. Experimental results show that proposed method outperforms other compared methods.

关 键 词：自商图像 局部二值模式 光照变化 人脸识别 直方图融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]