Image authentication using relation measures of wavelet coefficients

Abstract

In this paper, a novel watermarking technique for image authentication is proposed. This technique operates in the wavelet domain and makes use of semifragile watermarks for achieving better robustness. The image is first transformed into the wavelet domain and the relations between neighboring coefficients are utilized to authenticate the integrity of the image content. By carefully designing the processes of feature extraction and watermark embedding, the proposed method is able to identify malicious changes made on the image, while very insensitive to common image processing. In addition, the system is secure because not only the watermark is protected by a private key but also building relations between coefficients significantly discourages an adversary from performing the collage attack.