基于多方法融合的人脸图像光照纠正算法  

Illumination correction algorithm for face images based on combination of several algorithms

作　　者：郭宇聪[1] 张星明[1] 詹皇源[1] 张咏梅[1] 

机构地区：[1]华南理工大学计算机学院,广东广州510640

出　　处：《计算机工程与设计》2006年第9期1547-1549,1559,共4页Computer Engineering and Design

基　　金：广州市科技计划基金项目(2002XGP16);广州市天河区科技计划基金项目(2002XGP11)

摘　　要：提出了一种人脸识别图像的光照纠正预处理算法。该算法首先将均衡化引入仿射变换算法和光照补偿算法中,改进它们的处理效果;然后采用一种基于直方图灰度重心分布统计的方法,将多种预处理算法有效地融合在一起,得到新的光照纠正预处理算法。实验表明该算法具有光照无关性,能取得良好的光照纠正效果,并可大幅度提高复杂光照环境下的人脸识别率。A new preprocess algorithm was presented to correct the illumination of face images. In the algorithm, histogram equalization was introduced into AT algorithm and ICR algorithm to improve the performances of them. And method based on the statistic of distribution information of histograms＇ gray barycenter, was also adopted to combine several preproeess algorithms effectively. Experiment shows that the new algorithm is illumination invariant, which can achieve better preprocess result and can increase recognition rate of face images greatly in complex illumination environment.

关 键 词：人脸识别 预处理 仿射变换算法 光照补偿算法 光照无关 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]