Blind image-adaptive watermarking

Abstract

In this paper a new blind image-adaptive watermarking technique is proposed. The main contributions in this work are the following. First, a new spatial mask taking into account the Human Visual System (HVS) properties, is proposed. The mask is constructed based on the local variance of the cover image prediction error sequence. Second, an improved detection scheme has been developed, which is blind, in the sense that no knowledge concerning the cover image is required. The similarity measure used in the detector is the normalized correlation between the reproduced watermark and the prediction error of the watermarked and possibly attacked image (instead of the image itself). Due to the above modifications the proposed technique exhibits superior performance as compared to the conventional HVS-based blind adaptive watermarking. This performance improvement has been justified theoretically and verified through extensive simulations. In particular, the proposed technique is robust to additive white noise, JPEG and Wavelet compression, filtering etc.