一种基于类间类内双权重图像分块PCA的人脸识别技术  

A Human Face Recognition Technology of Block PCA Based on Between-class and Within-class Double Weight Images

作　　者：施志刚[1] 

机构地区：[1]南通航运职业技术学院管理信息系,江苏南通226010

出　　处：《杭州师范大学学报（自然科学版）》2014年第1期89-93,共5页Journal of Hangzhou Normal University(Natural Science Edition)

摘　　要：提出一种将加权分块图像和主成分分析(PCA)相结合的人脸识别方法.该方法首先根据同类训练样本的平均图像与所有训练样本平均图像的距离以及类内训练样本图像与该类平均图像的距离,分别定义类间和类内图像加权函数,以获得每个训练样本图像的权重;然后将训练样本图像分块,构建所有同位置加权分块图像空间;接着基于新的样本空间对所有同位置图像分别采用PCA方法提取特征;最后用最近邻分类器实现模式分类.实验结果显示该方法较普通MPCA方法有效提高了识别率.The paper proposed a method for human face recognition combining Principal Component Analysis （PCA） and weighted block images for enhancing the differences of between-class and within-class images .According to the distances between the average images of the similar training samples and the average images of all training samples ,as well as the distances between the within-class training sample images and the average images of the same class ,the weight functions of the between-class and within-class images were defined respectively to obtain the weights of every training sample images . And the training sample images were divided into block images to construct the weighted block images space in the same position .Then PCA was used to extract features from the block images in the same position respectively based on the new sample space .Finally ,the nearest neighbor classifier was used for realizing the pattern classification .The results indicate that this method can raise the recognition rate effectively compared with ordinary MPCA method .

关 键 词：加权分块图像 类间 类内 主成分分析 同位置 提取特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]