[Insight-users] Weighted normalized cross-correlation on ITK?

[Miguel Sotaquirá]

Hi everyone,

I'm new to ITK.

I'm implementing a block matching algorithm by using a template and a
floating image; this block matching algorithm makes use of weighted
normalized cross-correlation, which works pretty much the same as a
conventional normalized cross-correlation with the exception that for each
location of the template both template and the corresponding region of the
floating image are multiplied by a gaussian kernel (of the same size as the
template) thus giving a higher weight to central elements than peripheral
ones.

I've already implemented this algorithm in Matlab and I was wondering how
can this be done in ITK. Looking at the documentation and the examples I
found these approaches:

- itkNormalizedCorrelationImageFilter (
http://www.itk.org/Wiki/ITK/Examples/Images/NormalizedCorrelationImageFilter):
this filter allows for the computation of NCC.

- itkMaskedFFTNormalizedCorrelationImageFilter: as stated on the
documentation it "Calculates masked normalized cross correlation using
FFTs". The filter requires one fixedImage and one movingImage with a
corresponding pair of fixedMask and movingMask masks. However,each pair of
image-mask must have the same size (contrary to weighted NCC, where the
same gaussian mask is used for computation)

So I think I'll try to use itkNormalizedCorrelationImageFilter. My question
is: can weighted NCC be computed using this class? Can I use the method
SetMaskImage to compute the weighted correlation?

Thanks for any suggestions.

Miguel Sotaquira
PhD student
Biomedical Eng. Department, Politecnico di Milano
Milan (IT)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20120317/64a38c20/attachment.htm>