A Noninvertible SVD Technique for Watermarking Video Sequences

Abstract

Video watermarking has been proposed in recent years to solve the problem of illegal manipulation, copyright protection and distribution of digital video. Video Watermarking provides a means of forensic analysis for combating media piracy. In this paper, we present an effective, robust and imperceptible simulation of a video watermarking algorithm that is based on the algebraic transform of Singular Value Decomposition (SVD) where the watermark-bit information are embedded in the SVD-transformed video in a diagonal-wise fashion. The current implementation of this technology provides a forensic solution, in which each media provider marks each copy of a piece of content with a watermark specific to the consumer of the content (e.g. a distributor, a theater, etc.). In this case the provider can track pirated content back to the initial consumer. The paper also discusses the use of genetic and neural networks to be employed to optimize the watermarking requirements.