[Insight-users] different interpolators

Luis Ibanez luis.ibanez at kitware.com
Fri May 27 12:52:35 EDT 2005



Hi Yu,

When you use a BSpline interpolator, your Moving image is
implicitly smoothed by using a kernel of 4x4x4 pixels.

When you use a Linear interpolator, your Moving image is
implicitly smoothed by using a kernel of 2x2x2 pixels.

If your moving image is particularly noisy, it is to expect
that the BSpline interpolator will give better registration
results.



Easy way to verify/refute this guess:


    Run a Gaussian smoothing filter on the moving
    image, and then use this smoothed moving image
    as the new moving image in your registration.
    If you use a Gaussian sigma that is about 2 pixels
    wide, the result of the Linear interpolator should
    be closer to the Bspline.

    Note that the SmoothingRecursiveGaussianImageFilter
    expects Sigma to be specified in millimeters, not
    in pixels, so... take a look at your pixel size.



Just for the record,

You may want to be more careful when inferring generic
statments out of particular experiences.

The fact that *one* registration experiment is performing
poorly with a particular interpolator y by far not enough
for concluding that that interpolator cannot produce good
results. Every registration problem has between 10 and 30
parameters that must be fine-tunned. Different sets of
parameters will produce different results.

An Algorithm without its parameters
is an incomplete entitiy.




   Universal statements are almost always false.




  Regards,


      Luis




---------------------------------------------------------------
Yu Qi wrote:

> I use linear interpolater and BSpline(3rd order) interpolator 
> separately in my application of registration.
>  
> The result of using BSpline interpolator is very good, I can see it from 
> the resampled image.
>  
> But the result from Linear interpolator is not good. it has a larger 
> versor value and 10mm different along Z axis. The resampled image also 
> shows so.
>  
> I try to tune the parameters with linear interpolator, I got almost same 
> results.
> Does it means sometimes the linear interpolator can not get a good result?
> 
> 
> *Best Regards!*
> ** 
> *Yu Qi*
>  
>  
> 
> 
> ------------------------------------------------------------------------
> *Do You Yahoo!?*
> 150万曲MP3疯狂搜,带您闯入音乐殿堂 <http://music.yisou.com>
> 美女明星应有尽有,搜遍美图、艳图和酷图 <http://image.yisou.com>
> 1G就是1000兆,雅虎电邮自助扩容! 
> <http://cn.rd.yahoo.com/mail_cn/tag/1g/*http://cn.mail.yahoo.com/event/mail_1g/> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> 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