[Insight-users] Question: Evaluating the bias of Fourier Spectrum

[Luis Ibanez]

Hi Jiang,

At this point your best option is to calibrate a discriminant
using your own population of curves.

That is:

1) Select typical curves that you consider to be "Smooth"

2) Select typical curves that you consider to be "Not Smooth"

3) Use the current ITK example for computing the spectra of
    both populations.

4) That should give you two typical spectra, one for smooth
    curves, another for noisy curves.

5) Take those two spectra and use them as vectors in an M
    dimensional space, compute their orthogonal base and
    use the orthogonal base in order to discriminate the
    Spectrum of any new curve that you want to classify
    as smooth or not smooth.

(M) is the number of frequencies in your spectrum.




Regards,


   Luis




---------------------
Jiang wrote:

> Dear ITK users,
> 
> Maybe this question is not so appropriate to put in this maillist. 
> However I do believe that there are some cool guys
> 
> in this list who can answer my question.
> 
> The assumption is that if the curve is smooth, then its Fourier 
> Transform will
> 
> have high value in low frequency bands and low value in high frequency 
> bands. I want to use this method to measure
> 
> the smoothness of one curve. And I get the Fourier Transform of one 
> curve by the itk class.
> 
> My question is: is there a good method that can evaluate the bias of the 
> Fourier spectrum and give a quantitive result
> 
> for this issue?
> 
> I’m looking forward your suggestions and hints. Thanks a lot!
> 
>  
> 
>  
> 
>  
> 
> Jiang
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users