Digital watermarking in a perceptually normalized domain

One of the fundamental issues in digital watermarking is to find the best trade-off between imperceptibility and robustness to signal processing. Properties of human visual systems (HVS) have been exploited to achieve this goal. Previously, researchers have made use of frequency sensitivity, luminance sensitivity and visual masking effect of the HVS to adaptively control the amount of watermark energy to be embedded into different transform coefficients of the image. We present an alternative approach that first nonlinearly transforms the original signal to a perceptually uniform domain, then embeds a constant amount of watermark to each sample in this domain. We show that the derivation of an optimal watermark detector is straightforward in this approach, as opposed to some previous approaches. The advantage of exploiting the neighborhood-masking effect is also briefly discussed.