A New Near-Duplicate Detection System Using Object Semantics for Filtering Image Spam

As image spam becomes widespread and does a lot of harm, it is more important to filtering such spam effectively for now. Low accuracy rate and poor recognition of obfuscation techniques are main defects of current anti-image-spam technologies. For solving those problems, this paper proposes a new image spam near-duplicate detection system, applying image object semantics extraction to filter spam. The experimental results show that the system has better filtering performance than previous systems, particularly on higher accuracy rate and better anti-obfuscation effect, and effectively solves the above-mentioned problems.