结合对比度信息与LBP的分块人脸识别  

Recognition of intersected face based on contrast information and local binary pattern

作　　者：曹红根[1] 袁宝华[1] 朱辉生[2] 

机构地区：[1]南京理工大学泰州科技学院计算机科学与技术系,江苏泰州225300 [2]复旦大学计算机科学与技术学院,上海200433

出　　处：《山东大学学报（工学版）》2012年第4期29-34,59,共7页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助项目(61101197/F010402)

摘　　要：基于局部二值模式(local binary pattern,LBP)的特征提取方法忽略了图像灰度值变化的强度。针对这一问题,提出了一种结合对比度信息和LBP进行人脸识别的方法。首先采用LBP算子、VAR方差(variance,VAR)算子分别提取分块人脸灰度图像的LBP直方图序列(local binary pattern histogram sequence,LBPHS)和VAR直方图序列(variance histogram sequence,VARHS),然后将LBPHS和VARHS串联成LBP/VARHS,最后根据最近邻原则进行人脸识别。该算法能够提取有效的人脸纹理信息,而且能够大幅度地降低训练数据量,并且数据量的维数与原始图像大小无关。在ORL和YALE标准人脸数据库上的实验表明,该方法应用于人脸识别中,具有较高的识别率。The method of feature extraction based on local binary pattern（LBP） ignored the magnitude of the gray level differences. To solve this problem, a face recognition method based on contrast information and LBP was presented. First, LBP operator was used to extract the LBP histogram sequence （LBPHS） from block gray-level face images, and VAR operator was used to extract the VAR histogram sequence （VARHS）. Second, the above two operator were con- catenated into a single histogram sequence （LBP/VARHS）. Finally, the face recognition was realized based on the nea- rest neighbor principle. The proposed method could effectively extract the face feature, and greatly reduce the amount of training data. The dimension of the amount of data had nothing to do with the original image size. Experimental results showed that the proposed method could obtain better recognition rate on both ORL and YALE face database.

关 键 词：局部二值模式 对比度信息 直方图序列 特征提取 人脸识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]