基于人脸轮廓的人脸归一化方法  

Hybrid Eyes Location and Face Contour Normalization Method for Face Recognition

作　　者：王倩[1,2] 刘竞杰[2] 鲁魁魁[3] 

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001 [2]安徽工贸职业技术学院计算机系,安徽淮南232007 [3]中国科学技术大学计算机科学与技术系,合肥230027

出　　处：《合肥学院学报（自然科学版）》2013年第2期33-37,共5页Journal of Hefei University :Natural Sciences

基　　金：传感技术联合国家重点实验室基金(SKT1206)资助

摘　　要：人脸识别作为生物识别技术的一种,具有无接触、安全和方便的特点.人脸识别技术广泛应用于人机交互、交易认证及安防等领域,一直是生物识别技术研究的热点.人脸检测、特征定位、人脸归一化和特征提取是人脸识别研究的重点,决定着人脸识别系统的最终性能.设计了基于人脸轮廓的人脸归一化方法,根据归一化中出现的问题,进一步提出了人眼位置与人脸轮廓结合的人脸归一化方法.实验结果表明在Yale人脸图像库上人眼位置与人脸轮廓结合的人脸归一化方法具有更高的正确率.Comparing with other biological recognition methods, face recognition has the advantages of non-touching, safety and convenience. Face recognition has been widely used in human computer in- teraction, online transaction authentication, intelligent protection alert and etc. Face recognition Can be divided into several stages, including face detection, feature location, face alignment, feature extraction and feature classification. These aspects determine the performance of face recognition. In this paper, we present face normalization method based on the face profile and joint eyes location and face contour with face normalization method. The experimental results show that hybrid eye location and face con- tour normalization method has higher accuracy under the Yale data set.

关 键 词：人脸识别 特征提取 人脸归一化 ASM算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]