基于动态轮廓的彩色多人脸检测  

Multi-face Detection Based on Color Images Using Snake Algorithm

作　　者：尹红梅 朱再新 

机构地区：[1]扬州市出入境检验检疫局,江苏扬州225003 [2]扬州市科委,江苏扬州225003

出　　处：《电视技术》2007年第B08期134-136,142,共4页Video Engineering

摘　　要：提出了一种复杂背景下的人脸轮廓提取算法,算法包括以下步骤:先基于肤色分割图像,确定肤色区域的边缘;然后,利用人脸模板投影去除不可能是人脸的区域;再用Snake算法获取平滑轮廓;再用惯量矩得到拟合椭圆;最后利用简单的人脸模板对人脸区域进行确认。其算法的主要目的是解决人脸轮廓边缘点的不连续性问题,提取精确的人脸轮廓用于人脸分割和人脸识别,在一定程度上去除了部分遮挡造成的假轮廓边缘点,实验结果证明了算法的有效性。In this paper, an algorithm is put forward to acquire the contour of faces against complex backgrounds. Firstly, based on the result of skin-color segment, edges of skin-color region are determined. Secondly, for each region, adaptive searching schemes are designed to smooth the contour. Thirdly, ellipse fitting is made based on the moment of inertia. Finally, based on the result of curve fitting, face region can be verified accurately. The scheme is designed to resolve the discontinuity of face contour points, and acquire accurate contour to segment and recognize faces. False edge points caused by illumination and shutoff are wiped off to some extent. The experimental result confirms that the algorithm is efficient.

关 键 词：肤色模型 人脸分割 曲线拟合 惯性矩 动态轮廓 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]