[Insight-developers] Re: FW: [Insight-users] GradientImageFilter vs GradientRecursiveGauss
ianImageFilter
Joshua Cates
cates@sci.utah.edu
Tue, 30 Apr 2002 16:27:22 -0600 (MDT)
Hi,
I see the confusion now. Since DerivativeOperator is being used in
GradientImageFilter as a convolution kernel, it should be flipped across
all of its axes before sweeping through the image and taking inner
products. This will make the difference in sign for a first derivative.
I seem to have neglected to add a method in NeighborhoodOperator to do
this axis flipping, which would be very handy; so I'll add one and fix
GradientImageFilter tomorrow.
Josh.
______________________________
Josh Cates
School of Computer Science
University of Utah
Email: cates@sci.utah.edu
Phone: (801) 587-7697
URL: www.cs.utk.edu/~cates
On Tue, 30 Apr 2002, Miller, James V (Research) wrote:
> Josh,
>
> The GradientImageFilter uses the DerivativeOperator. Is this filter using the
> DerivativeOperator incorrectly? Or are the coefficient signs flipped?
>
> Jim
>
> -----Original Message-----
> From: Pieter Vos [mailto:pieterv@nat-239.azn.nl]
> Sent: Thursday, April 25, 2002 5:55 AM
> To: insight-users@public.kitware.com
> Subject: [Insight-users] GradientImageFilter vs
> GradientRecursiveGaussianImageFilter
>
>
> Hello,
>
> I compaired the outputs of the filter GradientImageFilter and the
> GradientRecursiveGaussianImageFilter.
> Both filters had the same 3D image: a black image with a white spot in the
> middle. The vectoroutputs however, where not the same:
>
> The vectors of the GradientImageFilter where pointing away form the spot.
>
> ------------------
> |
> | ^
> | |
> | <- ->
> | <- ->
> | |
> |
> |
> ------------------
>
> And the vectors of the GradientRecursiveGaussian where pointing to the
> white spot:
>
> ------------------
> |
> |
> | |
> | -> <-
> | -> <-
> | ^
> | |
> |
> ------------------
>
> The second picture is as how it should be.
>
> So, am I forgetting some basic theory (and should I study again)
> or should the outcome of the GradientImageFilter not be the opposite?
>
> Please help.
>
> Greetings,
>
> --
> Pieter Vos
> UMC Radboud Nijmegen
> The Netherlands
> mailto:pieterv@radiology.azn.nl
>
> _______________________________________________
> Insight-users mailing list
> Insight-users@public.kitware.com
> http://public.kitware.com/mailman/listinfo/insight-users
>