"Robust video watermarking based on the notion of interests region"
Computing System engineering
National School of Engineers (
M. Chokri BEN AMAR
HDR, University of Sfax Tunisia
ET
M. Henri NICOLAS
Professor, University of Bordeaux 1 France
The fast development of the means of communication and storage of data multimedia not only permits to recopy comfortably and without deterioration this type of documents, but also to modify them, to rebroadcast them or even to appropriate them without the author's agreement. The problem of the protection of the rights of the authors arises then in a crucial manner. In this context, the watermarking of the documents multimedia appears like a technological solution permitting the identification of the owner of a multimedia document. In the domain of the numeric video, the basis principle of the watermarking consists in introducing in the sequence of picture an invisible mark, but capable to be recovered with the help of a key, what permits the authentification of the author of the document then. To be efficient, these marks must not not only be destroyed by a phase of compression video, but also not of the algorithmic treatments specifically intended to destroy them.
The topic of this thesis has for objective the development of new approaches permitting to introduce with robust manner marks (or watermarks) in a video. It will first of all be about selecting the zones of the most suitable pictures to the introduction of the watermark. For it, several criterias should be taken in account. First of all, when a mark is introduced in a zone of the picture, it is necessary to introduce the same mark in the same place on all pictures of the sequence. Indeed, in the contrary case, a simple temporal filtering in the sense of the motion permits to destroy the mark. It will be also important to detect the less important zones visually in which of the modifications could be introduced more easily without they are visible. It is notably about the visible zones in few of different pictures, or of the zones or the human eye is little susceptible to detect the differences, as the zones of shades in movement. Not to arrive to u detection efficient of such regions, a phase of analysis permitting the creation of pictures mosaics notably (or sprite in the MPEG-4 context) and the detection of the shades in movement should be achieved. Finally, the marks themselves will be adapted to the features of the regions considered, in order to minimize the visible distortion, and no merely the simple report signal/noise.
The reliability of the methods developed will be tested while valuing their robustness facing phases decompression video, as well as to the principals types of algorithmic attacks aiming to destroy the watermark.
Key words : video watermarking, interests region
Références :
H. Nicolas. "New Methods for Dynamic Mosaicing". IEEE Transactions on Image Processing, Vol. 10, No. 8, pp. 1239-1251, August 2001.
G. Doerr and J.L. Dugelay. "New intra-video collusion attack using mosaicing". Proc of VCIP, Lugano, Switzerlan, July 2003.
G. Le Guelvouit and S. Pateux. "Wide spread spectrum watermarking with side information and interference cancellation”. Proc. SPIE, Santa Clara , CA , January 2003.