[ITK-users] Using MattesMutualMetric in V4 registration framework.

[Vishwa]

Hi All,
I am trying to register two 3d images, both OCT's from different patients.
Currently, I do the translation to roughly overlap the two volumes of
interest, then do the affine transform in 3 levels, and finally do the
BSpline transform based registration in just one level.

My Volume size is 200 x 200 x 600. I am using  a machine with 64GB of
memory and 8 core processor.

Am using Conjugate gradient line search optimizer for the Affine
registration with MMIv4 metric, and using LBFGSBv4 optimizer for the
BSpline registration, again using MMIv4 metric. The Bspline grid is 5 x 5 x
5.

It takes about 12 hours for the process to complete. Is this normal?

Also, in the previous MMIv3 version, I could set the number samples used to
calculate the metric. I am not sure how to do that here or whether it even
matters in the v4 framework.

I see that there is something like setusefixedsamplepointset function for
the metric, which is currently set to false. Does it mean it is using all
the voxels in the metric calculation?

I appreciate anyone helping me out here.

Best regards,
Vishwa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20140710/9a8f867e/attachment.html>