[Insight-users] Mattes Mutual Information and 2D/3D registration

[Tom Radcliffe]

I've stepped around the memory problems I was having with
multi-resolution registration by going to a single resolution, and that
seems to have worked.

I've now moved on to a much classier problem :-)

I'd like to use Mattes mutual information
(itk::MattesMutualInformationImageToImageMetric) and am running into the
problem that the DRR samples from itk::RayCastInterpolateImageFunction
all fall outside of the allowed range (m_MovingImageTrueMin to
m_MovingImageTrueMax in the metric class.)

As near as I can tell, this is a more general bug:  the histogram range
ought to be computed on the interpolated values in all cases.  For the
general ray casting case the fix is not obvious, because as the
transform changes the values may change a lot.  For the case I'm
interested in, it isn't a huge problem because I've got a pretty good
starting point for the registration I care about.  A few samples may be
lost as the transform changes, but I don't care.

So what I'm interested in first-off is a confirmation that my diagnosis
of the problem is correct:  the moving image histogram ranges ought to
be computed based on the interpolated values.  I'm using ITK 1.8.1, and
have queried the bug database and not found anything on this issue.

Cheers,
Tom

Tom Radcliffe, Ph.D., P.Eng.
Chief Scientific Officer
Improved Outcomes Software Inc.
613-549-6666 x4224


--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.6.11 - Release Date: 2005-01-12