Mutual Information Based Watermarking Detection in Wavelet Domain for Copyright Protection

A new robust watermarking approach is proposed in wavelet domain in this paper. It is robust to both compressive and geometric attacks. The watermark is embedded into the wavelet transform modulus maxima (WTMM) of the host image, which is shift-invariant compared to the other wavelet domain watermarking approaches. On the other hand, scale and rotation invariants are achieved by the geometric normalization during watermark detection. Mutual Information approach is proposed to improve the robustness of watermark detection, which can capture higher order statistics of image than that of the correlation detection. Case studies involve various attacks such as shifting, lossy compression, scaling, rotation and median filtering on the watermark are presented and discussed on their effectiveness.