基于BIF和KR-RCA的跨年龄人脸图像识别

Cross Age Face Image Recognition Based on BIF and KR-RCA

跨年龄人脸识别受到越来越广泛的关注。除了基因、人脸老化过程还受到生活方式等多种因素的影响。采用BIF和KR-RCA实现跨年龄人脸识别。首先对人脸图像进行旋转、对齐、剪切等归一化处理,然后对处理好的图像提取Mean-BIF特征,最后采用KR-RCA进行分类。经过在FG-NET和MORPG两个数据库上实验证明,提出的方法可以有效地实现跨年龄人脸识别。

Age-invariant face recognition has received increasing attention in the face recognition community due to its great help for government application(e.g.passport verification).Inaddition to biological factors,face aging processing is affected by many factors(e.g.lifestyle).Biologically-inspired feature and kernel regularized relevant component analysis are used for age-invariant face recognition.Firstly,the mug shot is preprocessed with rotation,alignment and cropping.Then,biologically-inspired feature(BIF)with mean pooling is extracted form the face image after preprocessing.At last,kernel regularized relevant component analysis(KR-RCA)is used to get the final identification.Experimental results on FG-NET and MORPH datasets,which are most widely used in face aging problems,are presented to show the efficiency of proposed method.