Re: Comparing/combining matrices (images)?

From: Tim Brauch (RnEeMwOs.pVoEst_at_tbrauch.cNOoSPAMm)
Date: 11/27/04 

Date: Sat, 27 Nov 2004 05:40:31 GMT

"QQ" <quick@quest.ion> wrote in news:lradnSjWnu-XvjvcRVn-iA@rcn.net:

>
> However, the values of the two matrices are on totally different
> orders of magnitude. For example, the values in matrix A range roughly
> from 1 to 1000. The values in matrix B range roughly from .0001 to .1.
>
> I would like to create a new matrix (matrix C) that emphasizes the
> similarities in both these matrices, specifically the absolute
> minimums (valleys). Seems to me that simply multiplying each value
> (matrixC[i][j] = matrixA[i][j] * matrixB[i][j]) and identifying the
> minimum would seem to be an obvious solution. Is this the best method,
> or is there a more robust, scientific method worth investigating?
>

Something that came to mind would be to "normalize" the matrices in a
sense. One thing I do when I am working with matrices as digital images is
find the maximum value, say M, in the matrix. Then I divide each entry in
the matrix by M. This will make all of the values fall in the range [0,1].

Perhaps something like this will help you.

 - Tim 

-- 
Timothy M. Brauch
NSF Fellow
Department of Mathematics
University of Louisville
email is:
news (dot) post (at) tbrauch (dot) com