Human Visual System Models in Digital Image Watermarking

Abstract

In this paper some Human Visual System (HVS) models used in digital image watermarking are presented. Four different HVS models, which exploit various properties of human eye, are described. Two of them operate in transform domain of Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). HVS model in DCT domain consists of Just Noticeable Difference thresholds for corresponding DCT basis functions corrected by luminance sensitivity and self- or neighborhood contrast masking. HVS model in DWT domain is based on different HVS sensitivity in various DWT subbands. The third presented HVS model is composed of contrast thresholds as a function of spatial frequency and eye's eccentricity. We present also a way of combining these three basic models to get better tradeoff between conflicting requirements of digital watermarks. The fourth HVS model is based on noise visibility in an image and is described by so called Noise Visibility Function (NVF). The possible ways of exploiting of the described HVS models in digital image watermarking are also briefly discussed.