image fusion

Hanif G Khalak hkhalak at
Fri Oct 17 14:38:30 EST 1997

Francois Vandermeersch (fvanderm at wrote:

:          I am trying to create some software which visualizes the fusion
: of two images.  My software
: receives matched medical images and it visualizes the superposition of
: these images.  On the 
: screen, you can change the percentage of superposition of the images :

I think that 'superposition' has a different meaning than 'fusion'.
John Strout's suggestions about Photoshop's layering approach may be
a way to get a 'superposition' of the two images.

Fusion, as it is used in the image processing community, usually implies
'co-registration' of some sort.  That is taking one image (or average
of images) as the template, transforming the other(s) contrast/intensities
into the same morphology (that of the template).  This is often used 
in fusing MRI/CAT (structural) images with PET/fMRT (functional) images
in 2/3-D medical applications.

:           Ex. : taking two matched images A and B, in normal case, the
: result of the superposition
: shows on the screen, 50 % of the image A, 50 % of the image B.  You can
: change the % of 
: superposition.  You can visualize for example that the fusionned image
: is composed of 60% of A and 40% of B 
: or 80 % of B and 20 % of A, ...  This software must conserve the
: contrast,  color of the source 
: images.

I am not sure what you mean by X% of image A.  Is that X% of some measure
of the morphology/shape, or the rms intensity, or ??  Or all of the above?
The weighted average of two images (a single 'smoothed' compromise image)
is a bit different, I think, than what you want (or is it?).  But in any
case, how do you 'conserve' contrast/color - in the X% sense?

With 'co-registration', you have to decide what the template is going to be
(one image), and what you get as a result are N images that represent
transformed versions of the original images you wanted to co-register.
This is often useful when you have images of N patients, and you
want to do comparative analysis across patients, on collocated data.

Please understand - I am not criticizing your question.  I'm interested
in what you mean, and the answer.  :-)

: 	  I don't know of any techniques to do that.  If you have some
: information about these techniques,
: some litterature (or reference), some algorithms, please E-mail me.

Try Morphology Digest :

There is also a web page for SPM, a medical image registration and stat
analysis package in MATLAB, but I couldn't find it at this moment.

Good luck,
  Hanif Khalak,	 hkhalak at
  Institute for Computational Sciences & Informatics,
  George Mason University, Fairfax, VA 22030, USA

More information about the Bio-soft mailing list