[Insight-users] Re: itkmattesMutualInformationImagetoImageMetric

[Karthik Krishnan]

Grace,

Yes, As Kajetan mentioned, the Mattes metric needs to be minimized. This 
may sound counterintuitive since Mutual information defined as :

I(X,Y) = H(X) + H(Y) - H(X,Y)  has got to be positive,

Yet the mattes yeilds values that are negative. Read the last line of 
the GetValue() method in 
itkMattesMutualInformationImageToImageMetric.txx. It reads :
     return static_cast<MeasureType>( -1.0 * sum );
In other words the classical MI value is negated.

The other mutual information metrics though, need to be maximized. For 
instance, if you use itk::MutualInformationHistogramImageToImageMetric, 
you will have to maximize it.

HTH
-karthik

Kajetan Berlinger wrote:

> Hi Grace,
>
> does your optimizer maximize or minimize?
> Use optimizer->SetMinimize(true) to make sure.
>
> cu
> kaj
>
>
> Grace Chen wrote:
>
>> Hi,
>>
>>> From the ITK software guide, it seems to indicate that the metric 
>>> value should get smaller (more negative) as the iteration 
>>> increases.  However, that's not the case for me....
>>
>>
>> I am very confused.  I've posted many questions here but no 
>> answer.... Please help!!
>>
>> Thanx!!
>>
>>
>> Grace
>>
>>   ----- Original Message -----   From: Grace Chen   To: 
>> insight-users at itk.org   Sent: Tuesday, May 30, 2006 4:44 PM
>>   Subject: itkmattesMutualInformationImagetoImageMetric
>>
>>
>>   Hi there,
>>
>>   I am using itkmattesMutualInformationImagetoImageMetric, and I am 
>> wondering if the mutual information value approaching toward 0 (it's 
>> always negative), does that mean a better regirtration transformation?
>>
>>   Thanx!
>>
>>   Grace
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> 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
>