基于曲线拟合的人脸分割算法  

Face Segmentation Based on Curve Fitting

作　　者：毋立芳[1] 沈兰荪[1] 孔潇[1] 朱旭娟[1] 

机构地区：[1]北京工业大学信号与信息处理研究室,北京100022

出　　处：《计算机学报》2003年第7期893-896,F003,共5页Chinese Journal of Computers

基　　金：国家"八六三"高技术研究发展计划 ( 2 0 0 1AA1113 91);国家自然科学基金 ( 90 10 40 13 )资助

摘　　要：提出了一种复杂背景下的人脸分割算法 ,算法包括以下步骤 :首先基于人脸检测结果确定内外轮廓及搜索路径 ;然后 ,利用设计的自适应搜索算法 ,对于每一条搜索路径提取出真正的轮廓点 ;最后利用人脸轮廓的平滑性通过曲线拟合来补充不确定边缘点 ,从而完成人脸分割 .该文算法的主要目的是解决人脸轮廓边缘点的不连续性问题 ,提取真正轮廓边缘点用于人脸分割 ,在一定程度上去除部分遮挡造成的假轮廓边缘点 。In this paper, an algorithm is put forward to segment faces in complex backgrounds. Firstly, based on the result of face detection, a search range and searching paths are determined. Secondly, for each searching path, adaptive searching schemes are designed to extract the true contour point, and to discard false contour point. Thirdly, polynomial curve fitting is made based on the true contour points. Finally, based on the result of curve fitting, face region can be segmented accurately. Since the face contour is smooth, by curve fitting, the no contour point searching path can be complemented. The scheme is designed to resolve the discontinuity of face contour points, caused by illumination, different postures and so on. The experimental result confirms that author’s algorithms is efficient.

关 键 词：人脸分割算法 曲线拟合 自适应搜索算法 人脸轮廓 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]