[Insight-users] Question about All points mapped outside..

[Jonathan Wong]

Hi, I'm new to ITK and I am trying to understand how the
MultiResMIRegistration example application works. I managed to run samples
images from brainweb and I got decent results (They were oriented
correctly). However, I created some of my own vtkImages from microCT scans
of bone specimens. The specimens I was making them from had be thresholded
so they were binary. I made one image. And then I rotated the specimen and
made a second image. I also made a smoothed version of the original and
rotated that to make another two images. (I did this because I read
somewhere that when using Mattes, the moving image should be smoother) At
first I got the following error with my specimens:

Line: 193
Description: itk::ERROR: MutualInformationImageToImageMetric(01763328): All
the sampled point mapped to outside of the moving image

However, after changing the number of iterations and learning rate I did get
the program to run fee from errors. However the output would hardly be
called accurate for any combination of the 4 images I created.

So I had a a few questions.

What is a good metric to use on bone volumes?

What makes for a good metric in general?

How does one configure the param.file for the MultiResMIRegistration example
correctly without playing a guessing game to see whether or not errors
result? (Like how can one tell, when one will get the Line 193 Error?)

Here's a link to one of the vtkImages.

http://www.yousendit.com/transfer.php?action=download&ufid=7A676C6B77938FB4
http://www.yousendit.com/transfer.php?action=download&ufid=E9D4EECF2A3CD065
http://www.yousendit.com/transfer.php?action=download&ufid=BAC0A34D221379C8

Thanks
Jonathan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20060607/7fc5f895/attachment-0001.html