A Masking Model of HVS for Image Watermaking in the DCT Domain

A substantial amount of work has been devoted to understanding the human visual system (HVS) and applying this knowledge to image processing applications. In this paper, we present a novel watermarking approach for applications in copy protection and digital archives based on perceptual masking. In the proposed scheme, local properties of the image and the features of the human visual system in discrete cosine transform (DCT) domain are taken into consideration in order to optimize the watermark strength. Based on NVF (noise visibility function) of the cover image, the mask is of the same size, and we incorporate the luminance masking effect of the HVS in the masking image, since the human eye is less sensitive to change in regions with high brightness as well as in very dark regions compared to mid-grey regions, where the distortion is most noticeable. Experimental results demonstrate that the new embedding rule enhances the quality of the watermarked images both subjectively and objectively. A comparison is made between the results of our proposed algorithm with that use only the textured area to demonstrate the superior performance of our new algorithm, while not degrading the watermark detection performance.