[Insight-users] Registration convergence

Jonathan Orban jonathan.orban at uclouvain.be
Wed Dec 5 03:15:45 EST 2007


Hi Andrew, 

In the demons algorithm, no metric is really minimized in a global point
of view. This method is local: the deformation field in each point is
computed by looking at local gradients. In that way, it is an
approximation of the optimization of the SSD. For further reading, I
recommend: Xavier Pennec, Pascal Cachier, and Nicholas Ayache,
Understanding the "demon’s algorithm": 3d non-rigid registration by
gradient descent, In MICCAI, pages 597-605, 1999.

Therefore, the computed metric (SSD) gives you an insight on the way
your algorithm is working (or not) but should not be taken as "the
minimized metric". To some extent, and depending on your images, you
should allow the rise of the computed metric to avoid local minimas. 

GetRMSChange, gives you an insight on the number of changes that have
occurred during the last iteration. If the amount of change is small
then you can consider that your algorithm has stabilized and will not
move anymore.

Have fun with ITK,

Jonathan

On Tue, 2007-12-04 at 09:28 -0600, Godley, Andrew wrote:
> Hello,
> 
> I am using Demons to register CT scans.  I print out filter->GetMetric()
> and filter->GetRMSChange() through the observer.  I assumed that it is
> the Metric that should be minimised, but I read that it is the RMS.
> 
> Is the RMS what is being minimised?
> 
> Does the metric represent something like the summed intensity difference
> between pixels in the image?  Why wouldn't I want this minimised?
> 
> I run registrations where the Metric reaches a minimum and then starts
> to rise, whilst the RMS continues to decline. 
> 
> I have been very happy with ITK and have been using it for 3 months.  I
> had been able to avoid asking questions until now, as I can't get my
> current test images, one with a much larger rectum, registered.
> 
> Thank you,
> Andrew
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users



More information about the Insight-users mailing list