摘要
主成分分析方法(PCA)和局部二元模式算子(LBP)相融合的特征提取方法结合了PCA在提取全局特征方面的优势和LBP在提取局部纹理细节方面的优势,能够从人脸图像中提取出较好的用于支持向量机(SVM)进行人脸性别识别分类的特征。在提取图像的LBP特征时,对传统的LBP方法做了改进,采用级联的方法提取图像的LBP直方图特征。并将提取出来的LBP特征与PCA特征相结合用于SVM,实验结果表明,LBP和PCA相融合的特征较单独的PCA特征和LBP特征在性别识别上具有明显的优势。
The method which combined the characteristic of principal components analysis (PCA) with local binary pattern (LBP)'s combines the advantage in global features of PCA with the advantage in Details of local texture of LBP and could extract better characteristics from face image for support vector machine (SVM) to gender classification.Using Cascade method rather than traditional LBP method to extract LBP histogram characteristics of images and combine the features of LBP and PCA from extraction for SVM. The experimental results show that the characteristics combined PCA with LBP has obvious advantages than the PCA and LBP separately in Gender identification.
作者
李昆仑
王命延
LI Kun-lun, WANG Ming-yan (Department of Computer, Nanchang University, Nanchang 330031, China)
出处
《电脑知识与技术》
2009年第10期8023-8025,共3页
Computer Knowledge and Technology
关键词
纹理
性别分类
主成分分析
局部二元模式
支持向量机
texture
gender classification
principal components analysis
local binary pattern
support vector machine