Francois Cayre


Email : cayre (at) irisa (dot) fr
Adresse : IRISA - Temics
Campus Universitaire de Beaulieu
35042 Rennes Cedex - FRANCE
Tél : +33
Fax : +33
Assistante de projet : +33
(Huguette Béchu)


Our paper on watermarking security with Teddy Furon and Caroline Fontaine was awarded as the best paper of IWDW 2004 in Seoul, Korea.

Technical Program Committee, Member, IEEE ICIP'05, September, 11--14, 2005, Genoa, Italy.


In 2000 I received the Computer Science Engineering diploma (Image, Decision, Signal) from Université de Technologie de Compiègne and the DEA degree (M.Sc. Hons.) in System Control from the same university (advisor : Franck Davoine). As a result of a partnership, I received in 2003 the Ph.D from both Université catholique de Louvain (UCL Louvain-la-Neuve, Belgium) and École Nationale Supérieure des Télécommunications (ENST Paris, France). My Ph.D. advisors were Profs. B. Macq (UCL) and H. Maître (ENST).

My works essentially deal with multimedia watermarking : recovery after local non-linear transforms of images (M.Sc.) and watermarking of 3D meshes (Ph.D.). I am currently a post-doc fellow at INRIA/Rennes (Temics team) where I study watermarking security aspects of spread-spectrum based watermarking techniques.

I was involved in several national and european projects, including AQUAMARS (RNRT, new watermarking strategies), CERTIMARK (EU-IST, watermark benchmarking), BRIC (EU-IST, MPEG-4 video broadcasting over ATM/IP) and FABRIANO (ACI INRIA, security of watermarking techniques). I also spent a wonderfull 2 weeks internship in the GEOMETRICA team at INRIA/Sophia-Antipolis in 2003 within the TeleGeo project (ACI INRIA, geometry and communications).

Research Activities

I have been working on several aspects of watermarking applications, including copyright protection, authentication, steganography and security. As a side-effect of my Ph.D., I also developped some aspects of 3D mesh compression and transmission using spectral decomposition.

Copyright protection

In the field of image watermarking, a very common attack is to perform non-linear local transforms of the images (StirMark) so that watermark synchronization becomes impossible. During my Ms.C. I proposed a new idea to combat this attack, it is called soft watermarking : watermark carriers are no longer pseudo-random sequences, but smooth functions that can easily survive such transforms.

In the field of 3D mesh watermarking, we tried to combine both watermarking and compression approaches to provide a comprehensive framework for secure 3D meshes transmission. We first developped a spectral 3D geometry coder and we added a watermarking module that acts in the transform domain. Since geometry information comes in a progressive fashion, so do the hidden information.


Once a document is protected, one also wants to ensure its authentication. Fragile watermarking can be used towards such a goal for 3D meshes. I designed a spatial watermarking scheme that has a well-defined robustness class and good properties in terms of flexibility, hidden information localization and probability of false alarm.


Another application of data-hiding techniques is called steganography and aims at embedding a fairly big amount of information inside a document. It can be used for augmenting the capabilities of the document in an appropriate decoder using legacy representation formats. During my Ph.D., I developped a steganography method that is both very simple and secure.

Watermarking security

In our point of view, robustness is different from security. We currently study the security of spread-spectrum based watermarking techniques : the watermark is expressed in terms of a linear combination of secret carriers. Robustness is the ability of the watermark to glue to the document, whereas security can be expressed in the number of watermarked documents needed to reconstruct the secret carriers. In order to achieve this goal, we use a blind source separation (BSS) technique called independent component analysis (ICA).

Related works

I also contributed two other problems : watermark algorithms benchmarking and 3D geometry transmission.

Watermarking algorithms benchmarking

Within the CERTIMARK project, our problem was to design a benchmarking utility for image watermarking algorithms. The platform had to be flexible enough to address different kinds of attacks with various parameter ranges and so on. A brief description of our work can be found in :

3D geometry transmission

When conceiving our joint watermarking-compression scheme for 3D meshes, we remarked that spectral decomposition was not suitable for transmission, i.e.: the mesh reconstructed with few spectral coefficients did not look acceptable. To provide a better mesh appearance, we proposed the first lapped decomposition for 3D meshes. The results show a much more appealing reconstructed mesh, even at the beginning of the transmission.

Other publications (selected)

Book chapter

I wrote an introduction to various popular watermarking techniques for 3D meshes in :

International conferences

National conferences

Research report


Together with Teddy Furon, we added an Independent Component Analysis class to the IT++ library for performing blind source separation. It is based on FastICA. If you use results given by this Fastica software in an article for a scientific journal, conference proceedings or similar, please cite :

Like my works ?

Buy me a book !

Webmaster: Valid CSS! Valid XHTML IRISA
Last time modified: 2006-03-30