[WM]: MPEG Reencoding

["matthias hub" <matthiashub@sohu.com>]

Dear watermarkers,

we just implemented some spread-spectrum watermarking algorithm for MPEG video and run now some attacks. The first attack came to my
mind was the reencoding, means using a different encoder than the official MPEG simulation group standard encoder will produce a
different result. First I used the free TMPGEnc 2.5, then now the TMPGEnc 3.0 XPress for reencoding from MPEG-2 to MPEG-2. The
differences in the dct coefficients are really obvious after reencoding, even if the frame type, the size, the bitrate, .. is set
the same. And apparently the videos look really the same.

My question is, if someone has found a paper about this topic of "Reencoding impact to MPEG videos". I know, that is very special
for every encoder, but maybe there are some results. I think this attack is slightly different from the JPEG compression attack,
because different parameters than the compression ratio have been altered. So the question is maybe the "Similiariy of DCT blocks"
or, more general, the "Similiarity of FFT functions" in presence of similar visual perception. A visual model is usally not
applicable that easy, because of a different value of the dc coefficient and the prediction frames.

Long way to go. Help would be appreciated. Thanks!

Greets,
Matthias



______________________________________________________________________________

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