Data hiding in MPEG-2 bit stream domain

Abstract

Digital watermarking, or information hiding, refers to techniques for embedding data into other data. In this paper, we consider the case in which the embedding data are video sequences; in particular, we concentrate on the problem of digital watermarking in the bit stream domain. Applications are not limited to copyright protection and user identification. For example, one could think of using information hiding techniques in order to give a description of the scene that could be used for video indexing; other applications include subtitling, multi-lingual services, teletext, etc. In many of these, it is desirable for the data hiding scheme to be able to work in the bit stream domain, since it is likely that the original sequences are stored in compressed format. The main characteristics of digital watermarks are invisibility and robustness. For invisibility, we mean that the watermark must not impair the visual quality of the original video. For robustness, we mean that the embedded information should hardly be erased without destroying the original data, resisting against processing that could, intentionally or not, tamper with it. The purpose of this paper is to introduce a method able to embed a certain amount of bits per frame in MPEG-2 coded video sequences, acting directly in the bit-stream domain. These bits could be used for any purpose for which the offered bandwidth could be wide enough.