一种基于KL变换的椭圆模型肤色检测方法  

An Elliptical Model Based on KL Transform for Skin Color Detection

作　　者：高建坡[1] 王煜坚[1] 杨浩[1] 吴镇扬[1] 

机构地区：[1]东南大学无线电工程系,南京210096

出　　处：《电子与信息学报》2007年第7期1739-1743,共5页Journal of Electronics & Information Technology

基　　金：国家973计划(2002CB312102)资助课题

摘　　要：肤色检测是计算机视觉中的一个重要问题,该文提出了一种新的基于KL变换的椭圆模型肤色检测方法。算法首先对训练肤色样本进行了分布均匀化处理,然后以KL变换为基本工具最终得到了肤色分布区域的椭圆边界方程。方法简单直观。实践表明,该算法能完成对各种不同环境条件下所拍摄图像的肤色分割,效果理想,其性能明显优于常用的域值界定法和单高斯模型法。Skin color detection is an important problem in computer vision. A new elliptical model based on KL transform for skin color detection is proposed in this paper. As to the proposed algorithm, firstly, the training skin color samples are made to distribute uniformly, then the elliptical boundary equation of skin color region is attained by means of KL transform. This method proves to be simple and intuitive, The experimental results on the images taken under a wide range of different environment demonstrate its efficiency and its clear improvement over the method based on explicit skin region boundary and single Gaussian model.

关 键 词：人脸检测 肤色模型 肤色分割 KL变换 GAUSS模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]