ROBUST IMAGE WATERMARKING USING FEATURE BASED LOCAL INVARIANT REGIONS

International Journal of Innovative

Computing, Information and Control ICIC International c_{°2008 ISSN 1349-4198}

Volume 4, Number 8, August 2008 pp. 1977—1986

ROBUST IMAGE WATERMARKING USING FEATURE BASED

LOCAL INVARIANT REGIONS

Lei-Da Li and Bao-Long Guo

Institute of Intelligent Control and Image Engineering Xidian University

Xi’an 710071, P. R. China

[email protected]; [email protected]

Jeng-Shyang Pan

Department of Electronic Engineering National Kaohsiung University of Applied Sciences

Kaohsiung 807, Taiwan [email protected]

Received May 2007; revised November 2007

Abstract. This paper presents a novel robust image watermarking scheme for resisting geometric attacks and traditional signal processing attacks simultaneously. The proposed scheme consists of two main phases, the feature point based watermark synchronization and the Discrete Wavelet Transform (DWT) based watermark embedding/extraction. In watermark synchronization, the locally most stable feature points (LMSP) are extracted from the scale normalized image to generate some non-overlapped circular regions. These regions are invariant to image rotation and scaling, and watermark synchronization is achieved using these local invariant regions. In the second process, a new image water-marking scheme is proposed in DWT domain. The watermark is embedded into all the local invariant regions repeatedly, and it can be extracted blindly. DWT domain water-marking is robust to signal processing attacks and watermark synchronization gets rid of the effect of geometric attacks. The integration of the above two techniques in the pro-posed scheme guarantees its robustness to both signal processing attacks and geometric attacks. Experimental results have demonstrated the advantage of the proposed scheme. Keywords: Image watermarking, Geometric attack, Watermark synchronization, Image normalization, Feature point

1. Introduction. The past decade has seen an explosion in the use and distribution of

digital multimedia data. Meantime, we have witnessed a rise in copyright encroachment.

Digital watermarking is a promising way to protect the copyright of digital contents [1].

Extensive schemes have been proposed for image, video and digital vector map [2-4].

However, many of the existing image watermarking schemes are vulnerable to geometric

attacks, such as rotation, scaling and translation (RST). This problem is most pronounced

when the original image is not available in watermark detection [5]. Geometric attacks

can defeat many watermarking schemes in that they introduce synchronization errors

between the watermark and the cover image. As a result, the watermarking schemes

resistant against geometric distortions have been the subject of quite much research. The

existing schemes can be classified into [6]: non-blind scheme, invariant domain embedding

[7,8], template based synchronization [9,10] and feature-based synchronization [11-15].

Recently, feature-based synchronization has become an active research area. This kind

of scheme belongs to the second generation watermarking [11]. Bas et al. [12] extract

feature points from the original image and decompose the image into disjoint triangles