[Insight-users] Questions about apodization in the process of registration

[Luis Ibanez]

Hi Jay,

The ITK image metrics for which this effect is relevant are already
normalized by the number of pixels in the overlapping region.

Those are

     - MeanSquaresImageToImageMetric
     - NormalizedCorrelationImageToImageMetric

If you are concerned about other metrics being subject to such
discontinuities you can verify your hypotesis by simply exploring
a range of the parametric space in a translation transform.

That will give you metric plots like those shown on the SoftwareGuide,

      http://www.itk.org/ItkSoftwareGuide.pdf

from those Metric plots you can judge how relevant the eventual
discontinuities may be.


 Regards,


       Luis


------------------
Jay Li wrote:

> Dear Luis,
>  
> Thank you for your reply. Sorry that I was not able to feed back 
> promptly. I didn't express the question clearly in my previous email.
>  
> During the registration process, the metric will be calculated based 
> on the overlap of the fixed and moving image for each iteration. Since 
> the area of the overlap may change for each iteration, the number of 
> total pixels counted as the overlap may change, so the value of metric 
> may have discontinuity that might cause the failure in finding the 
> global extreme.
>  
> To avoid above, different weights could be given to pixels or voxels 
> within the overlap. For example, when calculating the metric, pixels 
> near the boundary are given lower weights, but those ones near the 
> center are given higher weights. I don't know if ITK has done this 
> kind of stuff. Thanks.
>  
> Regards,
>  
> Jay
>  
>  
>
> */Luis Ibanez <luis.ibanez at kitware.com>/* wrote:
>
>
>     Hi Jay,
>
>     By apodization do you refer to the attenuation of
>     the high frequencies in the spectral domain of the
>     images ?
>
>     I assume that you are concerned about this due to the
>     resampling of the images during the registration process
>     as well as after the transform has been extracted.
>
>     If that's the case, you don't really need to be too
>     concerned. It practice this is rarely a problem, unless
>     your images are actually populated with white noise,
>     or at least blue noise, which are the type of spectra
>     having large components in the high frequency range and
>     therefore requiring attenuation in order to avoid aliasing
>     during the interpolation used for resampling.
>
>     For typical medical images, the high frequency components
>     are very low, and you are usually safe by using simple
>     interpolations. That is, the Fourier transform of the
>     shape functions used by typical interpolation are already
>     attenuatting high frequencies.
>
>     You may want to look at the documentation of the
>     SincWindowed interpolator where some of this issues
>     are discussed.
>
>
>     From the practical point of view your best options
>     for avoiding local minima are :
>
>     - Multi-resolution
>     - Good initialization
>
>
>
>     Please let us know if you have further questions.
>
>
>
>     Regards,
>
>
>
>     Luis
>
>
>
>     -----------------
>     Jay Li wrote:
>
>     > Dear all,
>     >
>     > I'm new to ITK. Could anyone give me a hint if ITK has
>     > taken care of the apodization related problems in the
>     > process of image registration. I looked through the
>     > manual and searched the online archives, but couldn't
>     > find anything related. Will there be any problems to
>     > find the global extreme without the apodization?
>     > Thanks in advance.
>     >
>     > Jay
>     > *>
>     >
>
>
>
>     *
>
> * *
>
> ------------------------------------------------------------------------
> *Do you Yahoo!?
> Yahoo! Mail Address AutoComplete 
> <http://us.rd.yahoo.com/mail_us/taglines/aac/*http://promotions.yahoo.com/new_mail/static/ease.html> 
> - You start. We finish. *
>
>*
>*
>------------------------------------------------------------------------
>*
>_______________________________________________
>Insight-users mailing list
>Insight-users at itk.org
>http://www.itk.org/mailman/listinfo/insight-users
>*
>