Fuzzy-ART based adaptive digital watermarking scheme

Abstract

A novel transform domain digital watermarking scheme that uses visually meaningful binary image as watermark has been developed. The method embeds the watermark information adaptively with localized embedding strength according to the noise sensitivity level of the host image. Fuzzy adaptive resonance theory (Fuzzy-ART) classification is used to identify appropriate locations for watermark insertion and its control parameters add agility to the clustering results to thwart counterfeiting attacks. The scalability of visually recognizable watermark is exploited to devise a robust weighted recovery method with composite watermark. The proposed watermarking schemes can also be employed for oblivious detection. Unlike most oblivious watermarking schemes, our methods allow the use of visually meaningful image as watermark. For automation friendly verification, a normalized correlation metric that suits well with the statistical property of our methods is used. The experimental results show that the proposed techniques can survive several kinds of image processing attacks and the JPEG lossy compression.