Optimum detection of robust perceptual-model-based image-adaptive watermarks

Image-adaptive watermarking based on sophisticated human perceptual models is capable of embedding watermarks with maximum strengths while incurring no perceptual loss. Very strong robustness as well as high information capacity can be achieved using these schemes. In this paper, the optimum detector for the perceptual-model-based robust watermarking is constructed, and the performance analysis is investigated. The new detector asymptotically is most efficient for weak signals, and particularly it is the most powerful for the perceptual-model-constrained watermarks. The experiments results validate our theoretical analysis.