Re: Invisible watermarking
- From: Prof Craver <xcottcraver@xxxxxxxxx>
- Date: Thu, 28 Jul 2011 10:15:14 -0700 (PDT)
On Jul 27, 7:28 am, gordonb.mp...@xxxxxxxxxxx (Gordon Burditt) wrote:
If you set GAIN = 3, and Watermark[x][y] has a value of 0 or 1, I
can wipe out most of the watermark by setting the lower 3 bits to
zero. (You can tell it's different from the original, probably,
but not what the watermark is).
That's not true: you will remove about 5/8 of the watermark energy,
because when X is congruent to 5, 6 or 7, attack(X)!=attack(X+3). The
rest will survive, and as the number of pixels increases, so will the
detection probability for a fixed error rate.
In any case, watermarks are usually stronger than this (and usually
zero mean, BTW) and often embedded in a transformed domain that makes
the signal harder to see and thus harder to remove for a given level
of imperceptibility. The 1996 algorithm of Cox et. al. would almost
certainly survive this kind of operation.
Is there any watermark that can't be wiped out by the process of
printing a copy of it on a cheap inkjet printer, then scanning it
back in with the image at a slight angle?
It's funny you should ask that. The paper I cited above described a
watermark that survived exactly this operation. Again, that's like 15
years old, and techniques have improved since then.
Is there any method of watermarking that, given an unmodified
copyrighted movie, can use a semi-published algorithm (this means
they reveal it to courts and expert witnesses, but they don't
change the algorithm between court cases), and generate a key
file to find the watermark:
"Copyright © 2010 MPAA. Death to pirates!"
embedded in whatever movie someone got caught downloading (with at
least 99.9% probability)? We'll assume that a "movie" is at least
half an hour of video long, and it might be something taken with
a home video camera.
Sure, that's easy. There are a great many algorithms for this, too
many to cite.
- Prev by Date: Re: Invisible watermarking
- Next by Date: Re: implementation of one-time-pad with Mersenne Twister PRNG
- Previous by thread: Re: Invisible watermarking
- Next by thread: Re: Invisible watermarking