摘要
研究人脸识别优化问题,不同程度光照对人脸图像的采集具有不利影响,使图像中包含一些噪声信息,而当前人脸识别算法没有考虑不同程度光照对人脸图像的影响,仅在光照变化不大时,识别正确率高。为了解决光照条件对人脸识别不利影响,提高脸识别正确率,提出一种多尺度Retinex(MSR)和支持向量机(SVM)相结合的人脸识别算法(MSR-SVM)。MSR-SVM首先采用MSR对人脸图像进行预处理,消除光照变化的不利影响,然后采用PCA提取人脸图像特征,消除一些噪声信息,最后利用SVM分类算法对人脸图像进行分类。采用Yale人脸库对MSR-SVM算法进行仿真测试,仿真结果表明,改进方法可以消除光照变化对人脸识别不利影响,加快了人脸识别速度,提高了人脸识别正确率。
Face recognition in the field of pattern recognition is an important research topic,and it includes two key steps：the feature extraction and classifier design.In order to solve the adverse effects of illumination conditions on the face recognition,the paper proposed a face recognition algorithm（MSR-SVM） based on multi-scale Retinex（MSR） and support vector machine（SVM）.Firstly,MSR-SVM adopted MSR to preprocess face image,and eliminated the harmful effects of light conditions.Then the PCA extracted facial image features,and finally face image was classified by classification algorithm SVM.Yale face library was used to test MSR-SVM algorithm.Simulation experiment results show that the MSR-SVM can eliminate the influences of illumination condition to face recognition,speed up the face recognition rate and improve the face recognition accuracy.
出处
《计算机仿真》
CSCD
北大核心
2011年第12期296-299,共4页
Computer Simulation
关键词
人脸识别
光照影响
主成分分析
支持向量机
Face recognition
Illumination affect
PCA
Support vector machine（SVM）
