Enhancing Detection of Digital Watermark against Resampling Attacks by Analyzing Periodicity of Interpolation

Abstract

Digital watermarking is one of the most fundamental techniques to protect copyright of digital contents. However, many watermarking systems are still weak against resampling attacks. In this paper, we propose a reinforcement scheme to complement conventional image watermarking systems by analyzing the periodic property of the interpolation. We first classify type of resampling regardless of embedded watermark. Then, the resampled image is reversely recovered to detect the embedded watermark. Empirical evidence from a large database of test images with three watermark systems demonstrates robustness to geometrical distortions such as scaling and rotation regardless of the watermarking systems.