基于分块相位一致性的人脸识别算法  

Face recognition based on modular phase congruency

作　　者：唐彩虹[1] 江艳霞[1] 王娟[1] 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机工程与应用》2012年第36期201-204,237,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60874002;No.61074016);上海理工大学光电学院教师创新基金(No.GDCX-Y1110)

摘　　要：针对基于可见光的人脸图像的识别容易受光照和表情变化的影响,人脸的表情变化仅限于局部等问题,以及图像的相位一致性特征不受图像的亮度或对比度影响的特点,提出了一种基于分块相位一致性的人脸识别算法。该算法用log-gabor滤波器对图像进行滤波,利用相位一致性模型提取相位一致性特征图像;对每幅特征图像进行分块主元分析(PCA)处理;融合所有子图像的距离信息,采用最近邻分类器进行分类识别。实验证明该方法具有更好的识别性能。The recognition accuracy of visible images is variant to changes in illumination conditions and facial expressions, and variations in face images are confined to local regions, while the phase congruency technique used in this research employs the local energy model, which is invariant to changes in illumination and contrast conditions in images. A method of face recognition based on modular phase congruency is used in this paper. The face images are convolved with a bank of 2D logarithmic gabor filters, and the images of phase congruency are obtained by using the model of phase congruency; the extracted feature maps are dealt by modular PCA; considering all the sub-images distance information, the nearest distance classification is used to distinguish each image. Experimental results show that the proposed method has better recognition performance.

关 键 词：相位一致性特征 LOG-GABOR滤波 分块 主元分析 人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]