A Blind Image Watermarking Based on Dual Detector

This paper presents a blind image watermarking technique based on a novel detection scheme which contains two detectors aiming at the positive attack and negative attack, respectively. The analysis on signal processing attack indicates that there exist unreliable transformed coefficients in the attacked image which result in errors of the extraction of watermark data. The novel detection scheme removes the unreliable transformed coefficients and employs the remaining reliable coefficients for the extraction of embedded watermark. The experimental results indicate the proposed technique improves robustness significantly, as compared to the existing single detector scheme.