文档介绍：南京邮电大学
硕士学位论文
人脸检测算法的研究与实现
姓名:凌亮
申请学位级别:硕士
专业:电子与通信工程
指导教师:卢官明
2011-06
南京邮电大学全日制工程硕士研究生学位论文摘要
摘要
人脸检测是指按照一定的策略将人脸从图像或视频背景中检测出来。由于其在安全访问
控制、基于内容的图像检索、视频会议、表情识别等众多领域具有重要的应用价值,近年来
人脸检测成为计算机视觉与模式识别领域非常活跃的研究课题。在阅读大量文献的基础上,
本文对人脸检测技术的现状与发展进行了概括,分析和总结了人脸检测的相关方法。
本文的主要研究工作包括以下几个方面:
(1)研究了一种基于 YCbCr 彩色空间的肤色检测方法。在比较肤色在各种色彩空间中的
聚类性后,利用在亮度和色度分离的 YCbCr 色彩空间对肤色进行分割,然后再进行二值化处
理。对二值化处理后的图像进行滤波、形态学处理,以消除噪声的影响。
(2)研究了一种基于 Adaboost 算法的人脸检测方法。并对其遇到倾斜人脸时,容易出
现漏检的现象,增加了新的特征模板,增强了对倾斜人脸的检测效果。
(3)针对肤色检测对复杂背景下的图像误检率高和 Adaboost 算法对多姿态、多人脸图
像检测效果不理想的问题,提出基于改进的 Adaboost 算法和肤色检测结合的人脸检测方法。
在 Adaboost 检测器前端加入了肤色检测,实验结果表明达到了优化检测的目的。

关键词:人脸检测, YC bC r 色彩空间,肤色检测,Adaboost 算法
I
南京邮电大学全日制工程硕士研究生学位论文 Abstract
Abstract
Face detection means to detecte face from the image or video background in accordance with
certain strategy. Because of its important applications in security access control,content-based
retrieval,video conferencing,facial expression recognition and other fields,face detection have
begun to be a very active research topic on the field puter vision and pattern recognition in
recent years. Based on the extensive reading of literatures,this paper elaborates to the present
situation and development of face detection technology,analyzes and summarizes the face detection
of the related methods.
The study work of this paper mainly includes the following aspects:
(1) A method of skin detection algorithm based on the YCbCr color space have been studied.
paring with several color space,this paper uses brightness and color separation in the
YC bC r color space for skin color it is carried out with the method of
binarization processing. In order to eliminate the effect of noise,the paper uses the method of the
filtering and mathematieal morphology.
(2) A method of face detection algorithm based on Adaboost algorithm have been studied. For
the disadvantages of Adaboost algorithms: When the detector e