基于内容的图像型垃圾邮件过滤技术研究

摘要

提出了一种图像型垃圾邮件的过滤方法，该方法不依赖于附属图像的文字信息，而是直接提取图像本身的视觉特征，包括梯度直方图、颜色直方图和 LBP 特征。分析了支持向量机(SVM)算法，基于该算法实现了图像型垃圾邮件的过滤，实验结果表明，LBP特征的识别效果好于梯度直方图和颜色直方图特征。%The image-based spam filtering scheme was proposed, which does not depend on the text label of the image, however the visual feature is extracted directly, including the image gradient histogram, color histogram and LBP features in the scheme. The Support Vector Machine algorithm was analyzed. Based on this algorithm, the image spam filtering was implemented. The experimental results show that LBP is the best as compared with the image gradient histogram and color histogram.