Image Watermarking Method Resistant to Geometric Distortions*

Abstract

In this chapter, the authors propose a new watermarking algorithm using an Rotation, scale, and translation (RST)-invariant feature of images. They use the feature vector as a watermark, not as a key. The authors look at various geometric distortions and their implications in watermarking methods. They focus on two-dimensional images. With the advances in digital media technology and proliferation of Internet infrastructure, the modification and distribution of multimedia data became a simple task. Many watermarking methods have been reported to be resilient to geometric distortions. From the signal processing's viewpoint, a size change of an image means the change of the sampling grid. The watermarks are generated through an iterative feature modification and verification procedure. A bispectrum feature vector is used as the watermark and this watermark has strong resilience to RST attacks. This approach shows the potential in using a feature vector as a watermark.