[Insight-users] New Metric suggestions

[N.E. Mackenzie Mackay]

Hey Everyone,

	Right now I am trying to register a set of points on an image.  I want 
the end result of the registration to have the summation of of the 
pixel values in the image where the points lie to be as high as 
possible.  In other words: I want the points to be in the brightest 
spot of the image.

	My image is an ultrasound image and the points represent the surface 
of a the bone extracted from a CT scan.  I am assuming that the 
brightest points in my ultrasound is the surface.  Hence why I want my 
points to end up there.  What is happening is the points optimize to 
the at different locations... while close not what I want.

	It's a rigid registration so only rotations and translations are 
necessary.  As of now I have been changing between a mean squared and 
normalized correlation metric.  Both giving different results that I 
hoped.  Instead of going to the brightest points the registrations 
seems to optimize at the highest variance.  E.G. If I had an image that 
was half white and half black the points would go to the intersection 
of the black and white area.  Even if there was a very bright area in 
the white section.  This is of course a simplified version of what is 
going on.

	 I want the end result to be the points lying on the brightest pixel 
values possible.

	if anyone could suggest a metric that is already written in ITK or 
even something that I could right my self I would be more then happy to 
hear.

Thank you for your help,
Neilson