[vtkusers] Normalized cross correlation

[David Doria]

On Mon, Oct 18, 2010 at 10:03 AM, David Doria <daviddoria at gmail.com> wrote:
>
> Has anyone written a normalized cross correlation filter? This seems
> like a pretty standard thing to do that is missing from VTK.
>
> I started one here:
> http://github.com/daviddoria/vtkImageNormalizedCrossCorrelation
>
> if anyone wants to help.
>


I wrote the algorithm. Unfortunately... it is VERY slow. I haven't even been
able to test it at all because even with a small (11x11) kernel, it is
taking about a minute just to do ONE ROW of the correlation on a 400x400
image. I am clearly doing something very wrong with the vtkImageData things.
Would anyone mind taking a look to see if they can see a better way to do
some of the core functions (SubtractMean, for example). I have some
DeepCopy's, but I didn't see a way around them, as I needed to actually
modify the data while keeping it intact in the original image.
http://github.com/daviddoria/vtkImageNormalizedCrossCorrelation

Thanks,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20101020/861a3a23/attachment.htm>