An image spam detection method
Chun-Yuan Su, Day-Fann Shen, Guo-Shiang Lin · 2017
In this paper, an image spam detection method was proposed. The proposed method has several parts: key block extraction, feature extraction, multi-level spam classifier. Key block extraction is used to extract the important information from image spams. Since color is one of important visual information to identify produces for humans, it is measured as a feature. To deal with geometric transform, spatial information among key blocks is extracted as a feature. After feature extraction, the number of key blocks is used to find some candidate clusters of image spam as the first-level classification for raising the efficiency of the proposed system. In the second level, the dissimilarity of color and spatial features is measured to determine whether the input image is spam or not. Experimental results show that the proposed method can determine whether the input image is spam well.