Capacity Bounds and Rate-Distortion-Analysis in Dither Vector QIM Watermarking With JPEG and JPEG2000 Attack

Capacity of a digital watermarking scheme is the maximum bit rates that can be embedded in a designated host signal with certain constraints. When host signal and channel attack are both Gaussian, the watermarking capacity is known. Rate-distortion-robustness analysis is investigating the relationship between information rate, embedding distortion and counter-attack robustness. The latter is normally measured by error probability. In this work, we present a dither vector QIM embedding algorithm for image watermarking against JPEG and JPEG2000 attack channels. It is showed that this kind of channel could be approximately viewed as a Gaussian channel. Based on the general Gaussian channel capacity results, we give some bound results for typical images. The theoretical relationship between the rate, distortion, and robustness for this vector QIM scheme has been studied. Simulation results are presented to show the mutual effects of these parameters in the scheme.