[Insight-users] Amoeba produces a non-orthogonal rotation matrix

[Kevin H. Hobbs]

On Thu, 2007-05-03 at 09:52 -0400, Kevin H. Hobbs wrote:
> I've used the attached code in the past to register images based on the
> region where I expect them to overlap. Now it produces an error message
> and quits after the first registration step. 

It looks like a test was added in CVS since I last used this code to
check if the rotation matrix could be used.

> 
> 
> I set the initial step size around line 246. How should I set the step
> so that it won't produce an invalid matrix?

I changed from Rigid3DTransform to Euler3DTransform which avoids setting
the matrix from the optimizer and instead sets the angles. It should be
more efficient anyway with 6 instead of 12 parameters.

> Instead of producing an error and quiting should the optimizer be
> informed that it has stepped out of bounds with a value like NaN?


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://public.kitware.com/pipermail/insight-users/attachments/20070503/ca8c2fc1/attachment-0001.pgp