Re: [WM]: modelling impact of media changes on watermark/data ?

[Francois Cayre <cayre@no-log.org>]

Dear Microsymore, Watermarkers, 


> It doesn't matter what domain we are really embedded/detecting our
> watermark (spatial/transform/or block transform domain).

( It depends wether you want to address resync issues easily or not. ) 

> ** Have any of you come across studies/papers which deal with effect on
> data due to format changes (maybe different media) ? Well modelling the
> degradation of watermark/data due to format/media changes ?


( The following assumes no resync issue )

What you can (somewhat) easily model is the impact of quantization on
the transform coefficients you use both for watermarking and
compression. If you embed your watermark, say, into block DCT
coefficients (Jpeg), and you compress with Jpeg2K, then modelling of the
attack becomes harder. 

However, I would suggest you to build an ad-hoc model : take your host
vector (block-dct coeffs) and check to what SNR re-compression (with
J2K) degrades it. Then build accordingly a watermarking scheme that can
resist such an attack in terms of SNR. 


> ** What about embedding/detecting in different domains ?

Intuitively, it not seem necessarily a good idea. 

I am sure clever watermarkers will come with a clever answer for you. 

François CAYRE

8, rue George Sand
GSM
+33 6 74 82 05 36
35700 Rennes
Home
+33 2 99 63 90 56 (+fax)
Email : Perso / PGP
(preferred / PGP) - Work
Work
+33 2 99 84 74 05 (page)


______________________________________________________________________________

Watermarking Mailing List - http://www.watermarkingworld.org/ml.html
To unsubscribe send email to "majordomo@watermarkingworld.org" with
"unsubscribe watermarking YOURMAIL" in the body.
______________________________________________________________________________