利用颜色和纹理特征的图像过滤方法  

A Color and Texture-based Method for Nude Image Filtration

作　　者：徐欣欣[1] 袁华[1] 张凌[1] 

机构地区：[1]华南理工大学广东省计算机网络重点实验室,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2004年第z1期24-27,共4页Journal of South China University of Technology(Natural Science Edition)

基　　金：高等学校优秀青年教师教学科研奖励计划资助项目广东省科技攻关项目(2004A10302003)

摘　　要：随着互联网络和多媒体技术的发展,色情图像的传播越发猖獗.为了有效地防止色情图像的传播,本文中提出一种综合利用颜色和纹理特征的色情图像过滤方法.该方法根据色情图像的颜色和纹理特征,建立肤色模型和皮肤模型对图像进行检测识别.利用该方法对752幅色情图像和4000幅正常图像进行过滤实验,结果表明该方法具有较高的准确率.With the development of Internet and multimedia technology, the spread of nude images becomes more and more rampant. In order to solve the problem, a color and texture-based method for nude image filtration is proposed in this paper. In this method, a color model and a texture model are set up according to the characters of the color and texture of images, which are then used to detect nude images. The proposed method is finally applied to the filtration of 752 nude images and 4000 benign images. And the results verify the accuracy of the method.

关 键 词：图像过滤 颜色 纹理 HSV空间 粗细度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]