[Insight-developers] Re: Question on the recursive Gaussian

[Luis Ibanez]

Well,

Originally I was computing the gaussian at the end of the pipeline,
and I agree with you in that probably the derivative filter at the
end of the pipeline will be a better approach for being robust to noise

I only moved to the beginning for the convenience of doing casting at
the input. Remember the issue with having this filter behave ok no
matter what image type we feed in ?

The filter is now taking any image type, and the first filter in the
minipipeline is casting from Input type to InternalImageType, which
happesn to be 'float' pixel type.

----

So, We could simply add another type:

SmoothingFilterType1 that will do this casting

and let the other smoothing filters be of type

SmoothingfilterType = InternalImageType -> InternalImageType

That probably will be better...
I just took the shortcut at that time.

Would you like to do this changes ?
Otherwise, I can do this over the weekend


    Luis



---------------------------------------
Miller, James V (Research) wrote:
Luis,

I notice in the filters using the recursive Gaussians, you construct a 
mini-pipeline where the first stage calculates the smoothed directional 
derivative and the subsequent stages just smooth.

(I have no idea what the right answer is here) Is is better to have the 
directional derivative first or last in the mini-pipeline.  Just 
thinking about propagation of errors, etc....



*Jim Miller*
*/_____________________________________/*
/Visualization & Computer Vision//
/GE Research/
/Bldg. KW, Room C218B/
/P.O. Box 8, Schenectady NY 12301/

//_millerjv@research.ge.com <mailto:millerjv@research.ge.com>_/

/_james.miller@research.ge.com_/
/(518) 387-4005, Dial Comm: 8*833-4005, /
/Cell: (518) 505-7065, Fax: (518) 387-6981/