固定背景下人脸图像的特征提取  

On Feature Extraction of Face Image Under the Fixed Background

作　　者：文可进[1] 蔡学敬[1] 王丽芳[1] 

机构地区：[1]重庆大学城市科技学院电气信息学院,重庆永川402167

出　　处：《四川职业技术学院学报》2012年第1期153-156,共4页Journal of Sichuan Vocational and Technical College

摘　　要：对于复杂人脸模式、脸部特征的提取,是人脸自动识别技术的关键,本文对固定背景下人脸图像特征的定位和提取算法进行了分析.文中首先对彩色图像背景进行分割,得到二值化图像;然后采用边界搜索方法确定人脸外接矩形;最后在色度空间中,结合SUSAN角点检测方法,实现人眼、嘴角定位,完成了人脸图像特征的提取和识别.实验证明,该算法在降低运算区域的同时,降低了运算复杂度和其他干扰因素,并且在速度、效率、准确性方面均有良好的性能.Theextractionofcomplexfacepatternand facial featureisthekeyto automatic face recognition technology. This article analyzes face image feature location and extraction algorithm under thefixedbackground. In thispaper, we finishthefaceimagefeature extractionandrecognition through the followingprocedures： we firstly obtain two value images fromcolor image background seg- mentation. Then we use the boundary searchmethod to determine the external rectangle of face. Final~ ly, we realize human eye and mouth corner location based on SUSAN corner detectionmethod in the color space. We expect to reduce the computational complexity and other interference factors, at the same time, to reducetheoperational area. The experimental results also prove thatthiscomprehensive al- ~orithmhas ~oodnerformance on sneed, efficiencv and accuracy.

关 键 词：人脸识别 边界搜索 肤色分割 外接矩形 SUSAN角点检测 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]