[Insight-developers] Bug 3647

[Karthik Krishnan]

Miller, James V (GE, Research) wrote:

> Karthik,
>
> You checked some changes to ImageFunction and 
> LinearInterpolateImageFunction that I would like to discuss.
>
> Inside LinearInterpolateImageFunction you put in a call to 
> IsInsideBuffer(). 
>
I was worried about that, but given that there are several index 
calculations, multiplications for triliear interpolation etc, would 6 
inequality comparisons add much ?

> The use case to date has been that it is the calling function's 
> responsibility to only evaluate the function at a pixel that is inside 
> the image.
>
> Also, you are caching the BufferedRegion as an ivar of the 
> ImageFunction.  There does not seem to be a reason for this as the 
> BufferedRegion should be available via an inline call on the image 
> itself (which the ImageFunction has access to).
>
The class has access to the image but is that an inlined call ?  The 
only reason, I was caching it as an ivar was that it would be used every 
call to EvaluateAtContinuousIndex().

> Looking at the bug log, it looks like you were trying to address an 
> issue whereby the last slice of a resampled image would be blank.
>
Thanks. That was the exact issue I needed to address. We have been 
registering MR images with poor interslice resolution and very few 
slices to go with (8 slices) and discarding the border slices made it 
even thinner leaving very few features to guide the registration.

The problem seemed to be one of the last slices not being considered 
inside the buffer, because of a tranform with a near zero z-translation.

> Dirk Padfield has a fix for this particular problem.  The problem is 
> one of numerical precision, and samples that were supposed to be on 
> the last slice were numerically just a smidge past the last slice.  
> Dirk changed some of the calculations in the ResampleImageFilter to 
> use integral/fixed point arithmetic.  This fixed the problems he was 
> having with the ResampleImageFilter. I'll have a Dirk check in these 
> changes.
>
Thank you very much. If you are crunched for time checking in those 
changes etc, could you please mail the filter code.

> *Jim Miller*
> */_____________________________________/*
> /Visualization & Computer Vision
> GE Research
> Bldg. KW, Room C223
> 1 Research Circle, Schenectady NY 12309-1027
>
> _millerjv at research.ge.com <_//_mailto:millerjv at research.ge.com_//_>_/
> /(518) 387-4005, Dial Comm: 8*833-4005/
> /Cell: (518) 505-7065, Fax: (518) 387-6981/
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Insight-developers mailing list
>Insight-developers at itk.org
>http://www.itk.org/mailman/listinfo/insight-developers
>  
>