Optimum detection and decoding of multiplicative watermarks in DFT domain

Digital watermarking is a powerful technique to help protect the intellectual property right as well as the security of multimedia content. Multiplicative watermarks have strong robustness, and are well-suited for the copyright protection. We investigate the optimal detection and decoding of multiplicative watermarks in DFT domain. For a class of non-Gaussian distributions, we derive locally optimum detector structure and extend it to decoding multiple-bit message. In watermark detection, optimal thresholding for given false-alarm rates is derived. The optimal decoding of multiple-bit messages in data hiding is constructed using generalized maximum likelihood estimation. Theoretical results are verified using experiments.