基于图像相似性检测的图像垃圾邮件过滤方法  

Method of image-based spam filtering of e-mails based on image similarity detection

作　　者：张秋余[1] 李建建[1] 余冬梅[1] 董建设[1] 王静[1] 贺洋伟[1] 

机构地区：[1]兰州理工大学计算机与通信学院,甘肃兰州730050

出　　处：《兰州理工大学学报》2008年第4期105-108,共4页Journal of Lanzhou University of Technology

基　　金：甘肃省高校研究生导师科研基金(0703-07)

摘　　要：日益泛滥的图像垃圾邮件给互联网用户带来极大的不便,如何对其实施有效过滤成为亟待解决的问题.分析图像垃圾邮件过滤中的关键问题邮件图像的特征提取.利用垃圾邮件重复发送、内容高度相似的特点,提出一种过滤图像垃圾邮件的新方法:提取邮件图像的综合特征值,以此作为目标邮件图像与垃圾邮件图像样本库相似性度量的依据,通过检测其是否相似来实现垃圾邮件图像的过滤.实验中,提取邮件图像的颜色、纹理和形状3种底层特征,其描述方法分别为颜色矩、共生矩阵统计量和不变矩.结果表明该方法对图像垃圾邮件的召回率达到95%以上.The ever increasing volumes of image-based spare e-mails are bringing more annoyance to internet users, and how to filter it has become a pressing problem. The key problem of image-based spam filte- ring-image feature abstraction was analyzed, and then based on the characteristics of the spam being sent repeatedly and their contents being highly resemble to each other, a new kind of image-based spam filtering method was proposed where the comprehensive feature of mailed images was extracted and taken as a basis of similarity of the object mail image to the image in sparn samples database. The similarity examination then would implement the spam filtering. In the experiment, the color, texture and shape features were extracted as sublayer vision features and described with color moments, intergrowth matrix moment statis- tics, invariant moments. It was showed that this method had a recall of 95%.

关 键 词：图像垃圾邮件 邮件过滤 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]