[Insight-users] Testing Robustness of Registration

[Olivier Salvado]

hello Isabelle,

There is a big difference between the two approaches that you suggest (I may
be wrong because I don't know exactly the itk implementation). If you write
an image at a new location, you will have to interpolate. Since
interpolation is filtering, you will in fact filter the new image that you
want to register. Once you find the correct parameters you will interpolate
a second time, thus filtering twice. Whereas by starting with random
parameters you will not do the first filtering step, only the second.
	Results between the two might differ depending on the frequency
content of your image and the kind of interpolation used.

	Another thing you want to consider. is that in real life, even if
the two images are the same, the noise is different. You should thus add
different noise after generating two images before trying to register them,
otherwise you will register the noise as much as the image. Since the noise
correlation is very high for zero misregistration, you will get an
artifactual high maximum at zero on your similarity measure (x-correlation,
mutual information, ...) that will take over any maximum from the image
content.

for more details you may want to look at our recent paper on that subject:
salvado and wilson, "Removal of local and biased global maxima in
intensity-based registration", Medical Image Analysis April 2007, 11(2). 



Olivier

__________________________________________________ 
Olivier Salvado, PhD
Case Western Reserve University | Case Center for Imaging Research 
University Hospitals of Cleveland | Department of Radiology | Wearn B49 
11100 Euclid Av. | Cleveland, OH 44106 
Ph. (216) 983 3426 | Fax: (216) 844 4987

-----Original Message-----
From: insight-users-bounces+olivier.salvado=case.edu at itk.org
[mailto:insight-users-bounces+olivier.salvado=case.edu at itk.org] On Behalf Of
IsabelleNg
Sent: Friday, May 04, 2007 20:37
To: insight-users at itk.org
Subject: [Insight-users] Testing Robustness of Registration


ITK-users, 

I wish to test one of the registration algorithm by applying random
transformations to the moving image. This is often done in papers that
report registration results as tests of robustness and capture range. Is it
valid to perform these tests by simply initializing the transform with
random numbers? i..e by calling 

registration->SetInitialTransformationParameters( randomXform)? 

Or, do we actually need to physically write out the randomly misaligned
images and then feed back into the algorithm? 

How would results differ with these 2 approaches?

Thanks,
Isabelle


--
View this message in context:
http://www.nabble.com/Testing-Robustness-of-Registration-tf3694891.html#a103
32457
Sent from the ITK - Users mailing list archive at Nabble.com.

_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users