[Insight-users] Mutual Information Error

Jordan Soet jds at interchange.ubc.ca
Wed Jun 8 17:34:02 EDT 2005


I just thought I'd give an update on my long going saga with this error.

I have pretty much given up on using the affine transform, as I have 
gotten the VersorRigid3DTransform working quite well, however, as I said 
in my last message I wanted to try the OnePlusOneEvolutionaryOptimizer 
with it. So, I tried replacing the versor optimizer with the 
evolutionary optimizer and I got the same '0/10000 samples outside of 
the moving image buffer' error message that I've been getting. At first 
I thought that it was just because the transform would only work with 
the specific optimizer, however, for some reason I then tried removing 
the code setting the optimizer scales and it 'worked' (if you count 
running through the entire program working. However, it went through 
4000 iterations and gave an output of [0,0,0,0,0,0] on 2 different images.)

Anyways, after I couldn't figure out how to get it working properly I 
then tried using the PowellOptimizer, and it seemed to work fine, it 
gave me a reasonable registration. However, when I tried to set the 
optimizer scales it gave me that error again.

I then tried not setting the optimizer scales with the affine transform, 
but that didn't work (although it did change the number of spatial 
samples which in the moving image buffer from 0 to around 20). I also 
tried it with the Powell Optimizer and it still had the same problem 
(although it upped the spatial samples to around 600 )

So anyways, I guess I don't really have any new problems, it's just that 
seems kind of weird to me, am I not supposed to use the optimizer scales 
with those optimizers? (I know in the software guide it says that 
setting the scales is most useful with the gradient descent optimizers.) 
Or is there something else I'm not understanding about them?

Thanks,
Jordan

Luis Ibanez wrote:

> Hi Jordan,
>
> The most likely cause is that you are not providing a
> good initialization for the Affine Transform.
>
> You probably want to use the CenteredTransformInitializer
> as described in the ITK Software Guide
>
>       http://www.itk.org/ItkSoftwareGuide.pdf
>
>
>
> Regards,
>
>
>     Luis
>



More information about the Insight-users mailing list