[Insight-users] An overlap invariant mutual information metric

Raghavendra Chandrashekara rc3 at doc . ic . ac . uk
Wed, 19 Nov 2003 09:39:55 +0000


Jessica de Ryk wrote:

> Hi,
> I was hoping someone might beable to help me with a way to convert the 
> mutual information metric to a normalized MI metric which is 
> independant of the amout of overlap between the fixed and moving images.
>  
> The following paper explains the theory but I have not been able to 
> figure out how to implement it in itk.
> *C. Studholme, D.J. Hawkes, and D.L.G. Hill, \/An overlap invariant 
> entropy measure of 3d medical image alignment/," Pattern Recognition, 
> vol. 32, pp. 71-86, 1999.*
> ** 
> I have been using itk release 1.2, is there any new relavent 
> developments in 1.4?
>  
> Jess

Hi Jessica,

We have implemented a NMI metric class which is in ITK 1.4. The class is 
called itkNormalizedMutualInformationHistogramImageToImageMetric. There 
is an example on how to use it in:

Testing/Code/Algorithms/itkNormalizedMutualInformationHistogramImageToImageMetricTest.cxx

Please let me know how you get along.

Regards,

Raghavendra