基于颜色和边缘特征直方图的图像型垃圾邮件分类模型

Image spam classification model based on color and edge histogram statistics

提出了一种新型的图像分类识别方法,该方法不依赖于对图像内容文字信息的提取,而是直接采用图像的颜色信息和图形边缘特征来构造用于图像模式分类的统计模型。通过在公开数据集上的实验结果表明,提出的模型对图像型垃圾邮件具有良好的分类能力,分类性能优于现有相关方法。由于该方法对图像型垃圾邮件的分类准确率高,且不受图像文字识别干扰技术的影响,具有良好的应用前景。

This paper presented a novel approach for image spam classification task,which did not rely on text information contained in the images,but made use of the color and edge features that could be extracted from image files directly to construct the statistical model for pattern classification.Experimental results on real public datasets demonstrate good performance of the proposed model.Since the approach is immune to content obscuring techniques,it offers a promising alternative for practical application.