[Insight-users] who know the implementation of MI in ITK?

[Karthik Krishnan]

Which MI metric are you using Mattes or MutualInformationImageToImageMetric
?

In the very last line of the metric you will notice that it is negated. ie
the computed metric is H(x,y) - H(x) - H(y). Hence it is to be minimized
(maximizing the true definition of Mutual information).

The other metric (MutualInformationImageToImageMetric) will yield positive
values and needs to be maximized.

--
karthik

On Wed, Jan 14, 2009 at 5:08 PM, Dongfeng Han <handongfeng at gmail.com> wrote:

> hello
> I have read the guide book on page 452
>
> I(a,b) =H(a)+h(B)-H(a,b)
> because H(a,b)<=H(a)+h(b), so I(a,b)should be >=0;
> so , we should maximize the I(a,b) for registration.
> But  my output for registration two 3d volume is like below. (the
> second column is the MI value)
> At each iteraton is decrease and the value is non-positive.
> It conflict with my analysis.
> so anybody can give me some suggestion?
> thanks
> 0   0   -0.271443   0
> 0   0   -0.286633   0
> 0   0   -0.286633   0.00620487
> 1   0   -0.300877   0.00620487
> 1   0   -0.300877   0.00304002
> 2   0   -0.307195   0.00304002
> 2   0   -0.307195   0.00132559
> 3   0   -0.223647   0.00132559
> 3   0   -0.300672   0.00132559
> 3   0   -0.308016   0.00132559
> 3   0   -0.308016   0.00116934
> 4   0   -0.28022   0.00116934
> 4   0   -0.307456   0.00116934
> 4   0   -0.308148   0.00116934
> 4   0   -0.308148   0.00112118
> 5   0   -0.305431   0.00112118
> 5   0   -0.308338   0.00112118
> 5   0   -0.307334   0.00112118
> 5   0   -0.308314   0.00112118
> 5   0   -0.308314   0.00107307
> 6   0   -0.276599   0.00107307
> 6   0   -0.307406   0.00107307
> 6   0   -0.308268   0.00107307
> 6   0   -0.30831   0.00107307
> 6   0   -0.308314   0.00107307
> 6   0   -0.308314   0.00107307
> 6   0   -0.308314   0.00107307
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>



-- 
Karthik Krishnan
R&D Engineer,
Kitware Inc.
Ph: 518 371 3971 x119
Fax: 518 371 3971
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20090114/d389504b/attachment.htm>