A blind image watermarking using multiresolution visibility map

Abstract

A perceptual pyramid watermarking method is proposed. The key idea is to use the contrast sensitivity of the human visual
system (HVS) to determine “invisible” regions where watermark energy can be adjusted providing an invisible and robust watermark.
These invisibles regions are obtained by computing a “visibility map” at each level of the Gaussian pyramid (GP). The watermark
is weighted by the local contrast and a global scaling factor. The embedding process is performed by modifying the values
in some levels of the Laplacian Pyramid (LP) using the spread spectrum technique. Afterwards, the watermarked image can be
constructed from the levels of the LP. For watermark detection, a blind detection scheme using the threshold-correlation based
technique is proposed. Finally, the performances of the watermarking method are evaluated in terms of invisibility and robustness
using some quality metrics and different attacks of Stirmark such as Gaussian noise, low-pass filtering, Jpeg compression
and cropping. This evaluation is performed for the choice of some parameters of the watermarking system depending on performances
such as invisibility and robustness. The design of our watermarking technique can finally be formulated as an optimisation
problem where the objective is to guarantee a trade-off between invisibility and robustness.

KeywordsContrast–HVS–Laplacian Pyramid–Quality Metric–Watermarking