[Insight-users] use of itkKappaStatisticImageToImageMetric

[Miller, James V (GE, Research)]

If there is a GetDerivative() implemented, it was from an early attempt.

The math for the derivative in not straightforward since the metric is 
based on counts. So a pointwise derivative (non-finite difference) is not 
straightforward.

You could always do a finite difference approach but there is not much a
difference between a finite difference approach to the derivative and 
an optimizer that does not need derivatives.

Again, we have another formulation for the derivative of this metric, 
we just need to review it.

Jim

-----Original Message-----
From: Karthik Krishnan [mailto:Karthik.Krishnan at kitware.com]
Sent: Tuesday, February 07, 2006 2:32 PM
To: Miller, James V (GE, Research)
Cc: mhg at unizar.es; insight-users at itk.org
Subject: Re: [Insight-users] use of itkKappaStatisticImageToImageMetric


So is the GetDerivative() method that exists in the metric now 
non-functional ? Is it not possible to have one implemented via finite 
differences ?

Thanks
karthik

Miller, James V (GE, Research) wrote:

>The Kappa statistics image metric does not compute derivatives wrt to the 
>transform parameters.  We a researcher take a look at it but it has not 
>been completed.
>
>You will need to use an optimizer that does not rely on gradients.  The
>AmeobaOptimizer and the OnePlusOneEvolutionaryOptimizers are candidates.
>
>You will also need to use a NearestNeighbor interpolator.
>
>Jim
>
>-----Original Message-----
>From: insight-users-bounces+millerjv=crd.ge.com at itk.org
>[mailto:insight-users-bounces+millerjv=crd.ge.com at itk.org]On Behalf Of
>mhg at unizar.es
>Sent: Tuesday, February 07, 2006 5:38 AM
>To: insight-users at itk.org
>Subject: [Insight-users] use of itkKappaStatisticImageToImageMetric
>
>
>Hello,
>
>I am trying to implement a registration method for binary images that uses the
>Kappa Statistic Metric. I wrote a standard program with this metric and a
>gradient descent based optimizer. 
>
>When I run my program, it tells me that this metric does not provide metric
>derivatives, so I guess that I should use another optimizer and maybe another
>interpolator.
>
>Could some of you tell me the correct combination for this metric?
>
>I would like also to know if there is any place where I could find the correct
>combinations for metrics, interpolators and optimizerd implemented in itk.
>
>Thanks in advance, 
>Monica. 
>
>
>
>
>_______________________________________________
>Insight-users mailing list
>Insight-users at itk.org
>http://www.itk.org/mailman/listinfo/insight-users
>_______________________________________________
>Insight-users mailing list
>Insight-users at itk.org
>http://www.itk.org/mailman/listinfo/insight-users
>
>  
>