基于图像相似性检测的图像垃圾邮件过滤方法

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.

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

The ever increasing volumes of image-based spam 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 filtering-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 spam 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 statistics,invariant moments.It was showed that this method had a recal1 of 95%.