Feature Based Watermarking

Abstract

A digital media product must be protected with a watermark. This mark must be resilient yet transparent to legitimate users of the media. There are those who seek to cloud the ownership of digital media by tampering with the embedded marks. For example, a small rotation, scale change or cropping of an image can make it difficult, or impossible, for the legitimate owner to recover the watermark. This is called a geometric attack. The authors present a new paradigm for rendering any watermarking scheme resistant to such geometric attacks. This is done by means of a new image transform to a Rotation, Scaling, and Translation (RST) invariant domain based on ideas from shape theory. They also propose extensions of this technique to video watermarking. Finally, the authors provide an example of how these ‘shape based’ concepts can be extended to more general relational databases provided an abstract notion of shape is employed.