Image spam filtering based on features of text region
基于文本区域特征的图像型垃圾邮件过滤算法

Text region embedded in most of spam images usually contains some discriminative features. An effective method based on the features of text regions to identify spam image was proposed. Firstly, the algorithm extracted the features of text regions that included the number and acreage, saturation, the number of characters and colors contained in text regions. Secondly, the spam images was classified by a support vector machine classifier. The experiment on real world data shows that the proposed algorithm can identify 98.5% of spam images and the precision is more than 98%.