Forensic tracking watermarking for digital cinema

Abstract

Many pirated copies of films may be available on the Internet before their official release. Over ninety percent of these copies were made by recording films with camcorders in movie theaters, so called camcorder piracy or in-theater piracy. As a deterrent to such piracies, this thesis introduces a forensic tracking watermarking scheme for digital cinema applications. The proposed scheme enables identification of the pirate who illegally recorded films in the theater. The information about when and where the film is playing is embedded as a forensic mark. When a pirate tries to record a movie, geometric distortion occurs in the copy due to the position of the pirate relative to the screen. In order to resist geometric distortion, the watermarking method is designed to estimate the distortion and recover from it. Moreover, the position of the pirate in the theater is computed by analyzing the estimated geometric distortion. Consequently, the watermark detector can inform when and where the copy was recorded from the embedded watermark, and the position of the pirate at that time from the estimated geometric distortion. To achieve the performance of the proposed scheme, a local auto-correlation function (LACF) is introduced. It is motivated by the auto-correlation function (ACF) which has been adopted to synchronize watermarks due to its robustness against affine transform. Since the ACF is vulnerable to projective transform and the geometric distortion by camcorder piracy includes the projective transform, we develop the LACF to have robustness to projective distortion. Also, the mathematical models for estimating projective transform and position estimation are provided to enable the forensic tracking system. On various movie clips, experimental results show that the presented scheme is robust against projective distortion as well as camcorder capture. Also, it is sure that the proposed position estimating model can judge the seat in a theater with a mean absol...