An efficient approach to still image copy detection based on SVD and block partition for digital forensics

As an important component of passive media forensics (PMF), content-based copy detection (CBCD) is appropriate for identifying piracy of digital content and preventing falsified evidence. In this paper, we propose a novel and efficient CBCD scheme for identifying different altered versions of original images. We employ singular value decomposition (SVD) and block partition to improve the robustness and discriminability of image content identification. To examine the validity of proposed copy detection algorithm, 32 attacks including geometric transformation, signal processing and image manipulation, are applied to copyrighted images to generate large virtual altered images. Experimental results demonstrate the efficacy of the proposed copy detection scheme in withstanding various attacks.