基于加权分块稀疏表示的光照鲁棒性人脸识别  

Illumination-robust face recognition based on sparse representation of blocks-weighted

作　　者：范守科[1] 朱明[1] 

机构地区：[1]中国科学技术大学自动化系,合肥230027

出　　处：《计算机应用研究》2015年第5期1563-1567,1571,共6页Application Research of Computers

基　　金：中国科学院战略性先导科技专项基金资助项目(XDA06030900)

摘　　要：针对光照变化对人脸识别的效果带来严重影响,提出一种对人脸识别的光照变化具有鲁棒性的方法,即基于加权分块稀疏表示的人脸识别方法。该方法首先对人脸图像进行离散余弦变换(DCT),通过去除DCT系数的低频部分来移除光照变化分量。通过反离散余弦变换得到光照归一化后的人脸图像,将人脸图像分块,独立地对每个子块作基于稀疏表示的分类,并对每个子块的分类结果进行加权投票得出测试人脸图像的类别。在Yale B、extended-Yale B、CMU-PIE和FERET人脸库上进行实验,实验结果表明该方法适用于光照鲁棒的人脸识别。The objective of this work is to recognize faces under variations in illumination. Previous works have indicated that the variations in illumination can dramatically reduce the performance of face recognition. To this end,this paper proposed an efficient method for face recognition which was robust under variable illumination. First of all,it employed a discrete cosine transform（ DCT） in the logarithm domain to preprocess the images,removed the illumination variations by discarding an appropriate number of low-frequency DCT coefficients. Then,it partitioned a face image into several blocks,and classified the blocks using sparse representation-based classification respectively. At last,on the basis of classification result of a block,it determined the identity of a test image through voting the classification results of its blocks. Experimental results on the Yale B,extended-Yale B,CMU-PIE and FERET databases show that the proposed method can achieve excellent recognition rates,and it＇s an efficient method for illumination-robust face recognition.

关 键 词：人脸识别 光照归一化 稀疏表示 加权分块 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]