计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (18): 22-24.DOI: 10.3778/j.issn.1002-8331.2009.18.006

• 博士论坛 • 上一篇    下一篇

复杂背景和光照多变的人脸检测方法

李全彬1,2,孙巧榆3,刘锦高1,李 明1   

  1. 1.华东师范大学 电子科学技术系,上海 200062
    2.徐州师范大学 物理与电子工程学院,江苏 徐州 221116
    3.华东师范大学 计算机科学技术系,上海 200062
  • 收稿日期:2009-02-25 修回日期:2009-03-26 出版日期:2009-06-21 发布日期:2009-06-21
  • 通讯作者: 李全彬

Face detection in complicated backgrounds and different illumination conditions

LI Quan-bin1,2,SUN Qiao-yu3,LIU Jin-gao1,LI Ming1   

  1. 1.Department of Electronic Engineering,East China Normal University,Shanghai 200062,China
    2.College of Physics and Electronic Engineering,Xuzhou Normal University,Xuzhou,Jiangsu 221116,China
    3.Department of Computer Science and Technology,East China Normal University,Shanghai 200062,China
  • Received:2009-02-25 Revised:2009-03-26 Online:2009-06-21 Published:2009-06-21
  • Contact: LI Quan-bin

摘要: 复杂背景和光照多变的人脸检测是自动人脸识别系统需要解决的重要问题之一,首先利用YCbCr空间进行肤色分割,获取候选人脸区域,然后提取相应区块的LSMQT特征,并利用SNoW分类器进行特征分类,最终获取人脸在图像中的准确位置。实验证明,该方法对复杂背景和光照多变的图像中的人脸具有较高的正确检测率。

关键词: 人脸检测, YCbCr颜色空间, 连续均值量化变换, 稀疏网络模型

Abstract: Face detection in complicated backgrounds and different illumination conditions is one of the important problems to be solved in the automatic face recognition system.The image including faces is searched by using YCbCr color space firstly,then,the features of the segmented skin color areas are extracted through LSMQT method,further,the SNoW classifier is introduced.Finally,the features and classifier are combined for the task of face detection.The results of experiments show that this method has a high face detection correct rate to the images with complicated backgrounds and different illumination conditions.

Key words: face detection, YCbCr color space, Successive Mean Quantization Transform(SMQT), Sparse Network of Winnows(SNoW)