融合加权颜色相关图和改进LBP的彩色人脸图像识别  

Color Face Image Recognition of Combining the Weighted Color Correlogram with the Improved LBP Method

作　　者：叶继华[1] 陈亚慧[1] 胡蕾[1] 

机构地区：[1]江西师范大学计算机信息工程学院,南昌330022

出　　处：《小型微型计算机系统》2015年第12期2778-2783,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61462042;61262036)资助

摘　　要：彩色人脸图像识别的关键是如何获取较好的彩色人脸图像信息.根据现有算法的不足提出加权颜色相关图和LBPT算法,并采用有效的方法结合这两种方法实现高效的彩色人脸图像识别.加权颜色相关图不仅体现了彩色人脸图像中颜色的整体分布情况,还反映了颜色之间的空间相关性以及颜色的分布紧密度.LBPT算子是通过自适应获取邻域半径、确定邻域半径与邻域像素点个数关系以及阈值设定来反映灰度图像的纹理特征.二者结合表征了彩色人脸图像的颜色特征和纹理特征.实验证明,加权颜色相关图和LBPT算子的结合在识别精度上有一定的提高.The key to color facial image recognition is how to obtain better color face image information. In this paper, we proposed the weighted color correlogram and the LBPT algorithm according to the shortcomings of exiting algorithms ,and adopt effective method to combine the two methods to realize effective color face image recognition. In color face image,the weighted color eorrelogram not only indicates the holistic distribution of the color, but also reflects the spatial correlation among colors as well as the tightness of color distribution. The LBPT algorithm reflects the texture features of gray image through adaptively obtaining neighborhood radius, certai- ning the relationship of neighborhood radius and neighborhood pixel number, and threshold setting. The combination of the two charac- terized the color features and texture features of color face image. Experiments show that, the recognition accuracy has a certain im- provement when we combine the weighted color correlogram with the LBPT algorithm.

关 键 词：彩色图像 人脸识别 加权颜色相关图 LBPT算子 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]