Optimum detection of multiplicative watermarks using locally optimum decision rule

Multiplicative watermarks have very strong robustness, and they are well-suited for the copyright protection. In this paper, new detector structures for the optimum detection of multiplicative watermarks are derived. It is shown that the observations should be raised to the power of the shape parameter of the distribution before they are correlated with the watermark. For commonly used Gaussian distribution a quadratic correlator is obtained. A generalized linear correlator is also derived based on the Laplacian distribution. The performance analysis of the proposed detector is examined. The theoretical results are verified by the experiments.