类肤色背景中提取人脸轮廓的方法研究  

Studies on face contour extraction from approximate skin color backgrounds

作　　者：刘睿[1] 王晓东[2] 

机构地区：[1]西安邮电学院电子与信息工程系,陕西西安710061 [2]西安交通大学生命科学与技术学院,陕西西安710049

出　　处：《计算机应用》2005年第12期2855-2858,共4页journal of Computer Applications

摘　　要：提出了一个基于肤色并融合多种信息的人脸轮廓提取方法。首先在TSL色彩空间求取肤色概率图,选取种子点,然后利用多源信息进行区域生长,提取出人脸轮廓;为克服区域生长计算量大的缺点,采用了变分辨率图像金字塔策略。经实例验证,该算法能够快速准确地从类肤色背景中较好地提取出人脸轮廓,且具有较高的抗噪性和应用适应性。A new face contour extraction method was proposed by useing skin color and multiple information fusion. The method computed skin probability image based on TSL color space, then selected seed point, finally used multiple information to carry out region growing, extracted face contour. The method employed image pyramid process to avoid heavy computing burdens caused by region growth. Experiment results indicated that the proposed algorithm can extract face contour from approximate skin color backgrounds efficiently and accurately, and bears noise endurance and application adaptability.

关 键 词：人脸轮廓提取 肤色概率图 信息融合 图像金字塔处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]