[Insight-developers] GetMinimumMetricValuePosition

[Karthik Krishnan]

Rupert Brooks wrote:
> Hi folks,
>
> Im going to repeat a question I asked last month - I asked it just
> before ITK 3.8 release, so everyones minds were elsewhere.  Basically
> this involves a bug that was closed, but I think still exists.  I
> think that the ExhaustiveOptimizer does not agree with the definition
> of its base class (Optimizer) and so cannot be dropped in as a plug in
> replacement for another optimizer.
>
> Im curious if this is a design decision?
>   
Yes.

The optimizer to plot the metric space and exhaustively scan it to 
verify registrations. Ironically, the ExhaustiveOptimizer isn't an 
"optimizer" at all :)

The CurrentPosition at Iteration 't' is not a more optimal position than 
one at an earlier iteration. It is simply the position in metric space 
that we are currently scanning.

The MinimumMetricValuePosition is the best position so far found.

With the other optimizers, the current position is the best position so 
far found, hence this one needs another method not present in the base 
to get the optimal position.

Thanks
--
karthik

> Thanks
> Rupert
>
> On Mon, Jul 14, 2008 at 1:18 PM, Rupert Brooks <rupe.brooks at gmail.com> wrote:
>   
>> Hi,
>>
>> I was just examining the solution for the itkExhaustiveOptimizer
>> problem reported here
>> http://www.itk.org/Bug/view.php?id=5342
>>
>> One of the problems reported was that the optimizer could not be used
>> in the registration framework - its GetCurrentValue reports the last
>> place explored, which is not necessarily the best one.
>>
>> The current test for this optimizer uses the method
>> GetMinimumMetricValuePosition - and this works fine for finding the
>> minimum - so the test passes.  However, this does not seem to be a
>> general method for all optimizers.  Specifically - the
>> ImageRegistrationMethod uses ->GetCurrentPosition to find the optimal
>> transformation parameters.
>>
>> To my way of thinking the bug is not fixed - but the bug was closed.
>> Am i out in left field here?  Note that i was the original reporter of
>> the bug - so perhaps im not getting something.
>>
>> Cheers,
>> Rupert
>>
>> --
>> --------------------------------------------------------------
>> Rupert Brooks
>> McGill Centre for Intelligent Machines (www.cim.mcgill.ca)
>> Ph.D Student, Electrical and Computer Engineering
>> http://www.cyberus.ca/~rbrooks
>>
>>     
>
>
>
>   

-- 
Karthik Krishnan
R & D Engineer,
Kitware Inc,
Ph: +1 518 3713971 x119
Fax: +1 518 3714573