RE: [WM]: Pseudorandom matrix in Matlab

["Kim,H.J." <alwaysrainbow@hanmail.net>]

Dear Chuck:

As I mentioned before, my suggestion allows pseudo-randomness and
equal number of 1's and 0's. Pseudo-randomness can be obtained
if you use the same seed in Matlab. Help of RAND will give you how to
use the same seed.

There are randomness test procedures.
However, forget them and just rely on Matlab.

Another reliable and mathematically sound method is
using M-sequence. The other day, a watermarker posted a website
that contains the M-file for M-sequence. Make the M-sequence
to be 2-D. In order to make 8*8 matrix, use 8-register
LFSR. Then, you can get 63-bit sequence. It lacks one bit.
Patch 1 or 0 in order to make equal number of 1's and 0's.
Turn 0's into -1's.

You can make 8*8, 16*16, 32*32, ..., 2^n*2^n matrices.
However, it is difficult to generate 5*5, 6*6,7*7 matrices
with the LSFR. However, Matlab allows you to generate any size of matrix.
There would be nice theory to generate pseudo-random matrix.
Dr. Tirkel of Scientific Technology, Australia,
and his colleagues published an interesting paper 
(see http://www.physics.monash.edu.au/~ron/papers/ICPR98.PDF).

hope to be helpful,

khj


-----Original Message-----
From: Chuck Rewood [mailto:riwut@nii.itb.ac.id]
Sent: Thursday, July 26, 2001 12:38 PM
To: Kim, Hyoung Joong
Subject: RE: [WM]: Pseudorandom matrix in Matlab


Thanks, you are quite right, I haven't define what 'pseudo-randomness'
in my application,

Here I explain my situation (definition of my pseudo-randomness):
1. I should be able to create pseudo-random matrix, with equal number
   of 1's and -1's
2. The matrix itself must be dependent to the key (password) inputted
   by user(s), therefore the same key means the same pseudo-random
   matrix. (Think of translating hash function into pseudo-random
   [-1, 0, 1] matrix with exactly zero mean)
3. No mater how long is the key, matrix size is always the same (8x8 or
64x64)
   (just like creating fingerprint into pseudorandom matrix form)

I'll your idea, but another question is, how could I guarantee the
randomness ?

Any suggestion or other idea ?

Best regards,

Riwut Libinuko


-----Original Message-----
From: Kim, Hyoung Joong [mailto:khj@cc.kangwon.ac.kr]
Sent: Wednesday, July 25, 2001 3:13 PM
To: Chuck Rewood
Subject: Re: [WM]: Pseudorandom matrix in Matlab



What you want is pseudo-randomness and equal number of 1's and -1's?
Still, they need the exact definition of pseudo-randomness.
I think it's a good topic to do research.

Anyway, how about generating any sequence using RANDINT.
Change 0's into -1's. If 1's are more than -1's, and vice versa,
make the excessive 1's or -1's to be 0's.
Then, we have the same number of 1's and -1's and small number of 0's.

Does it make sense?

khj


=======================================
Prof. Hyoung Joong KIM
Dept Control/Instrumentation Engineering
Kangwon National University
Chunchon 200-701, Korea
khj@kangwon.ac.kr
http://multimedia.kangwon.ac.kr
=======================================





Prof. Kim, Hyoung Joong
Dept. Control/Instr. Eng.
Kangwon National University
Chunchon 200-701, Korea
khj@kangwon.ac.kr
http://multimedia.kangwon.ac.kr
+82 (17) 251-6343 (Mobile)
+82 (33) 250-6343 (Office)
+82 (2) 533-1719 (Home)

==================================================
¿ì¸® ÀÎÅͳÝ, Daum
Æò»ý ¾²´Â ¹«·á E-mail ÁÖ¼Ò ÇѸÞÀϳÝ
Áö±¸ÃÌ ÇÑ±Û °Ë»ö¼­ºñ½º Daum°Ë»ö
http://www.daum.net




______________________________________________________________________________

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