Image spam classification model based on color and edge histogram statistics
Weizhong Qian · Jisuanji yingyong yanjiu · 2010
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.