Robust Image Watermarking Against Geometric Attack

With the rapid development of the Internet and extensive use of multimediatechnology, the transform and sharing of information bring great convenience topeople’s work and life. How to protect the ownership of digital multimedia informationefficiently has become an important problem. Digital watermarking has been developedfor solving such problems and become a hot issue in multimedia security community.Most existing methods are considered to be robust against common image processing,such as compression, low filtering, etc. However, there are few methods efficientlyresolving geometric attacks, especially complex geometric attacks.This paper takes images as example and studies the robust watermarking algorithmagainst geometric attacks. This paper proposes two geometrically robust watermarkingalgorithms mainly using image feature extraction, muti-scale geometric analysis andimage statistics. The main contributions are listed as follows:First, a watermarking algorithm base on feature matching in Curvelet domain isproposed. This method restores the distorted image through calculating the geometrictransform parameters such as rotation, scaling, etc. The image restoration resolves theproblem of de-synchronization between the distorted image and the embeddedwatermark. Experiment results verify this algorithm is effective and robust againstcommon image processing and common geometric attacks.Second, a watermarking algorithm based on image restoration and the Zernikemoments of local area is proposed. This method restores the distorted image using atransform model which is optimized by RANSAC iteration and the method is veryuseful for the complex geometric attacks. The Zernike moments are employed todescribe the global characteristics of local invariant regions which are relativelyindependent to the location deviation and the numerical error in restoration image pixelvalues. Experiment results verify this algorithm is effective and robust against commonimage processing and complex geometric attacks.