[Insight-users] Fwd: co-occurrence matrix

[Zachary Pincus]

I'm forwarding this to the insight-users list in case anyone has a 
similar question or such a question in the future... The answer will 
follow momentarily.

Zach


Begin forwarded message:

From: Tanja Stumpf <t.stumpf at dkfz-heidelberg.de>
Date: June 30, 2004 4:20:50 AM PDT
To: Zachary Pincus <zpincus at stanford.edu>
Subject: Re: co-occurrence matrix

I wanted to use your newly written 
itkScalarImageToGreyLevelCooccurrenceMatrixGenerator class. And when I 
read about the submitted test I couldn't reconstruct how the given 
histogram was composed.  With the given test image

1 2 1 2 1
1 2 1 2 1
1 2 1 2 1
1 2 1 2 1
1 2 1 2 1
you observed with offset (0,1) and (1,0) you get the 
Cooccurrence-Matrix (only the interesting part):

    0 1 2
  +------
0| 0 0 0
1| 0 6 9
2| 0 9 12

  In my calculations I got the following "North"-Cooccurrence-Matrix:

N  0 1  2
  +-------
0| 0 0  0
1| 0 12 0
2| 0 0  8

and the "East"-Cooccurrence-Matrix (the occurances of (1,2) are also 
added to those of (2,1) to get a symmetric matrix):

E  0 1  2
  +--------
0| 0 0  0
1| 0 0  20
2| 0 20 0

finally you get this matrix:

    0 1  2
  +--------
0| 0 0  0
1| 0 12 20
2| 0 20 8

  Is there a mistake in my thoughts? Or are any their any specification 
I didn't see?  Thanks for your help!
  Tanja