Time-Variant Watermarking of MPEG-Compressed Digital Videos

Abstract

Watermarks allow one to embed information into digital videos in a way that is imperceptible to the viewer. This information can be used to establish ownership, trace origin of copies, and verify the integrity of the video. Watermarking may be compared to injecting additional energy; to ensure that this injection remains unnoticeable, it should be as small as possible. We outline an approach that permits a significant increase of the amount of information that can be accommodated in a watermark without any increase in the complexity of the process, namely time- variant watermarks. Since data compression is an important aspect in storing and distributing digital videos, we formulate our approach assuming the video is represented in an MPEG format. We discuss implementation issues of time-variant watermarks, with special emphasis on their advantages over the usual time- invariant watermarks. We comment on defeating attacks using filtering, cropping, resizing, and other standard methods used to defeat watermarks, such as changing existing frames, as well as new attacks, such as removing, repeating or permuting frames.