Digital image watermarking using independent component analysis

Abstract

In many watermarking applications, tracking the image copy ID is another requirement besides the ownership verification. Furthermore, it is a common demand to keep the original image from public accessibility during the watermark extraction process. In this paper, we propose a new watermarking method called WMica that bases on the Independent Component Analysis (ICA) technique. The proposed method employs a two-watermark embedding scheme; one watermark is to identify the ownership and the other serves as the ID for each copy of the original image. In the extraction scheme, an ICA algorithm is applied together with a down-sizing technique so that we can estimate all the watermarks without accessing the original image and the prior information about the watermarks. The new method, undergoing a variety of experiments, has shown its robustness against many salient attacks. It also exhibits a capability in image authentication. © Springer-Verlag 2004.