基于正面人脸图像的人脸轮廓的提取  

Extracting the Face Contour Line for Frontal Face Image

作　　者：向元平[1] 王国才[1] 乔汇东[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南省长沙市410075

出　　处：《微计算机信息》2007年第22期243-244,264,共3页Control & Automation

基　　金：国家自然科学基金(NO:60173041)<计算机网络分布式资源认证存取控制问题的研究>资助

摘　　要：近年来,人脸识别技术成为当前模式识别和人工智能领域的一个研究热点,人脸轮廓提取是人脸特征检测和人脸识别等人脸图像分析的重要前提,但至今仍没有得到圆满的解决。该文针对复杂背景的彩色人脸图像,提出并实现了一种人脸轮廓提取的方法。首先滤波和二值化人脸图像,在二值化图像中采用一种快速的肤色区域边界提取算法确定人脸区域;再在人脸区域内,采用轮廓跟踪技术提取出人脸轮廓。实验结果表明,该方法具有很高的精度和很强的鲁棒性。Recently, face recognition becomes a focus in the pattern recognition and artificial intelligence field. Extracting the contour llne for face image is important premise of face detection and face recognition. Aiming at the color image with complicated background, a novel method of close contour line extraction is presented. First, we filter and double-threshold the face image. In the dou- ble-threshold image, we use a fast extraction algorithm that based on the boundary of the color region to determine which the face region is. Then in the face region, we use the contour tracking to extract the contour line. Computer simulation shows that this algorithm has high accuracy and good robustness.

关 键 词：二值化 区域分割 轮廓跟踪 轮廓提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]