摘要
针对光照变化人脸识别问题,提出了一种基于局部归一化融合熵加权Gabor特征方法。首先,计算类熵加权向量;然后,对图像进行局部归一化,并且计算输入图像的Borda计数,从而消除低值Gabor jet比较矩阵;最后,利用投票得分策略和k近邻分类器完成人脸识别。在扩展YaleB和AR人脸数据库上进行实验评估,在AR上的识别率可接近100%,相比其它几种较新的局部特征方法,本文方法取得了更高的识别精度,并且在一定程度上减少了计算开销。
For the issue of face recognition with illumination variation,a fusion method based on local normalization and EW-Gabor features is proposed.Firstly,class entropy weighting vectors are calculated.Then,images are local normalized,Borda counts of inputting images are calculated so as to eliminating comparison matrix of low-value Gabor jet.Finally,vote score strategy and k neighbor classifier is used to finish face recognition.Proposed method is estimated by experiments on extended YaleB and AR databases,recognition accuracy of proposed method on AR can be close to 100%.Experimental results show that proposed method has higher recognition accuracy and less computation overhead than several advanced local feature methods.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第11期38-41,共4页
Laser Journal
关键词
局部归一化
GABOR特征
Borda计数阈值
人脸识别
光照变化
熵权法
LKey Llocal normalization
Gabor feature
Borda counting threshold
Fface recognition
Illumination variation
Entropy weighted method
