CAJ | 학술논문

为解决变光照下人脸识别率低问题，提出一种鲁棒性强的光照人脸处理算法（ISSR）。首先对亮度图像进行粗估计，然后通过非线性运算在反射图像中对粗估计的亮度图像进行补偿，最后采用支持向量机建立人脸识别分类器，并采用Yale、CMU-PIE和AR人脸库进行仿真测试。结果表明，相对于其它光照处理算法，ISSR算法增强了人脸图像的质量，提高不同光照条件下的人脸图像识别正确率，鲁棒性更强。
위해결변광조하인검식별솔저문제，제출일충로봉성강적광조인검처리산법（ISSR）。수선대량도도상진행조고계，연후통과비선성운산재반사도상중대조고계적량도도상진행보상，최후채용지지향량궤건립인검식별분류기，병채용Yale、CMU-PIE화AR인검고진행방진측시。결과표명，상대우기타광조처리산법，ISSR산법증강료인검도상적질량，제고불동광조조건하적인검도상식별정학솔，로봉성경강。
In order to improve the image recognition rate in illumination conditions, this paper proposes a good robust complex illumination face enhancement method based on improved single scale Retinex algorithm. Firstly, the illumination image approximately is estimated, and then compensates the approximate estimation in a nonlinear way, and finally, the face classifier is establish based on support vector machine, and three face database are used to test the performance. The results show that, compared with other methods, the proposed method has improved the quality of the image and can help improve the face recognition performance in complex illumination conditions.