[Insight-users] KullbackLeiblerCompareHistogramImageToImageMetric

[Karthik Krishnan]

You might want to look at:

1. The reference in the doxygen, where KullbackLeibler is used to 
register 2D-3D angiographs. (A. Chung, W. Wells, A. Norbash, E. Grimson, 
“Multimodal image registration by minimising Kullback-Liebler distance”, 
MICCAI, 2002.)
2. Kullback Liebler is used to register series of 3D-US to CT. 
(Multi-modal 3D registration using the Kullback-Leibler distance, 
Stippel, BIOMED 05.)

thanks
regards
kk

jef vdmb wrote:

> Hi,
>
> I think I understand now what the idea is, but I still have some 
> questions? Is the first joint histogram (acquired manually) 
> representative for MR- US registration in general?
>
> Say I have registered (manually) a 2D US and the 3D MR. Their joint 
> histogram can now be caculated.
>
> -Is this histogram then usefull for registrations of other images that 
> are located in a similar position of the fixed or moving image?
>
> - Can it be used for registration using the same moving or fixed 
> image, in the case that the other image is located in another position?
>
> - Would it be usefull (representative) to use for ANY MR-US 
> registration (with the same kind of MR and US images) no matter where 
> they are located?
>
>
> I hope I have made my doubts clear,
>
> Thank you very much for you swift reply Karthik,
>
> Jef
>
>
>> From: Karthik Krishnan <Karthik.Krishnan at kitware.com>
>> To: jef vdmb <jvdmb at hotmail.com>, ITK <insight-users at itk.org>
>> Subject: Re: [Insight-users] 
>> KullbackLeiblerCompareHistogramImageToImageMetric
>> Date: Wed, 20 Apr 2005 20:46:23 -0400
>>
>> Thanks for mentioning. The notation in the software guide is probably 
>> confusing.
>>
>> p_A(i) is intended to represent the joint histogram of the fixed 
>> image (say CT) and the first reference image (say US). Assume that 
>> you've somehow registered them (say manually). p_B(i) is intended to 
>> represent the joint histogram of the training image(s), say another 
>> US image which you wish to register to the fixed CT image just the 
>> same way the first reference image was registered. So the goal is to 
>> minimize the KullbackLeibler distance between the two joint 
>> histograms. I guess it will be clearer if the notation were changed 
>> from p_A(i) to p_ref(x,y) and p_B(i) to p_training(x,y').
>>
>> Thank you
>> kk
>>
>> jef vdmb wrote:
>>
>>> Hi,
>>>
>>> I am trying to find out which metric behaves the best for my multi 
>>> modal registration problem. The images are a 2D US and a 3D MR 
>>> (should you have any suggestions, based on their modality).
>>>
>>> I am plotting the metric value for all multi modal metrics 
>>> available, but I am having problems understanding how to use the 
>>> KullbackLeiblerCompareHistogramImageToImageMetric. In the 
>>> SoftwareGuide, it seems you only need your two images. When I go to 
>>> the Doxygen pages however, a "training histogram" is mentioned.
>>>
>>> Could you please clarify the use of this metric, any parameters that 
>>> need to be set (besides epsilon), or known preprocessing 
>>> improvements (like normalisation).
>>>
>>> Thank you,
>>>
>>> Jef
>>>
>>>
>>> _______________________________________________
>>> 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
>