基于自适应融合色度与亮度特征的彩色人脸识别算法  

A color face recognition algorithm based on adaptive fusion of chroma and luminosity features

作　　者：崔法毅[1] 

机构地区：[1]燕山大学河北省测试计量技术及仪器重点实验室,秦皇岛066004

出　　处：《高技术通讯》2015年第1期89-96,共8页Chinese High Technology Letters

基　　金：秦皇岛市科学技术研究与发展计划(2012021A004)资助项目

摘　　要：鉴于彩色人脸图像所包含的鉴别信息远多于灰度人脸图像,将色度马氏距离图引入彩色人脸识别中,提出了一种基于自适应融合色度与亮度特征的彩色人脸识别算法。该算法基于YDhDr颜色空间分离彩色人脸图像的色度与亮度信息,构建出基于色度信息的马氏距离图,同时分离出基于亮度信息的灰度图;通过自适应融合色度与亮度特征来构建彩色人脸识别特征;基于小波包结点能量的特征表示方法,分别在实数域和复数域中提取并融合色度与亮度分量的最崖鉴别特征向量,实现色度与亮度特征的最优化互补;使用基于方差相似度的分类器获得人脸识别结果。实验表明,该算法识别率高、鲁棒性好。Considering that color face images contain much more identification information than gray face images, the chroma Mahalanobis distance map was introduced into color face recognition, and a color face recognition algorithm based on the adaptive fusion of chroma and luminosity features was proposed. The algorithm separates the chroma and luminosity information of color face images based on the YDbDr color space to construct Mahalanobis distance maps based on chroma information, and extracts the luminosity information based gray maps from orignal color fa- cial images by separation; constructs the recognition features of color face images through the adaptive fusion of chroma and luminosity features ; constructs and fuses the optimum feature discrimintion vectors of chroma and lumi- nosity maps in the real domain and the complex domain respectively by using the feature representation method based on the energy of wavelet packet sub-nodes to achieve the optimal chroma feature-luminosity feature comple- mentation; and uses the classifier based on the variance similarity degree to obtain the face recognition results. The experimental results show that the proposed algorithm has the higher recognition rate and the better robustness.

关 键 词：彩色人脸识别 色度马氏距离图 YDbDr颜色空间 小波包变换 自适应特征 融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]