Watermark detection using channel estimation in the quantization based watermarking system

In this paper, we propose a novel blind detector for the quantization based watermarking systems. In literature, diversity and attack characterization based blind detectors are employed in these systems. These detectors both use the reference watermarks to characterize the attacks and the information watermarks to transmit the hidden data. The proposed detector is based on the channel estimation through reference watermarks. It estimates the reliabilities of the watermark channels and calculates the channel parameters by using the statististics of the reliabilities. Then, it determines the realiable watermark channels by using the proposed thresholding method. Finally, it combines the information watermarks extracted from the reliable watermark channels by properly weighting them with their corresponding channel parameter to recover the embedded information watermark. The simulation results demonstrate that the proposed blind detector is more robust than existing detectors against various channel distortions and malicious attacks.