[Insight-developers] A possible class submission

[Samson Timoner]

I have a new ImageToImageMetric that is very nearly ready to be submitted
to ITK. It is based on the Kullback-Leibler (KL) Distance. Essentially, you
have two training images that are aligned, and two test images you wish to
align. You try and make the histograms of the overlapping test images be
similar to the histogram of the overlapping training images. Similarity is
measured using the KL metric.

Because KL is only 1 way to compare 2 histograms, I've written the code as
2 classes. The first I call "CompareHistogramImageToImageMetric" which
inherits from HistogramImageToImageMetric.  This class is an abstract class
which forms a histogram from two training images. Thus, a sub-class must
simply define a way to compare the "Training" Histogram to the "Testing"
histogram.

I've also written KullbackLeiblerCompareHistogramImageToImageMetric. This
class only includes the method to compare Histograms using K-L.

I've written two test programs to go into Testing/Code/Algorithms, similar
to the tests in there for other ImageToImageMetrics.

Questions:

 1) Would you like me to submit the two classes and the test function?
 2) Other than the code I mentioned, do I need anything else to submit?
 3) Any Comments?

 Cheers,

-- Samson

P.S. The reference paper is from:

 Albert C.S. Chung, William M. Wells III, Alexander Norbash, and W. Eric L.
 Grimson, Multi-modal Image Registration by Minimising Kullback-Leibler
 Distance, In Medical Image Computing and Computer-Assisted Intervention -
 MICCAI 2002, LNCS 2489, pp. 525 - 532.