[Insight-users] Efficient way to calculate statistics in neighborhood?

[Luis Ibanez]

Hi Isabelle,


If you need to compute the statistics for all the
pixels of the image you may want to use the
StatisticsImageFilter.


If you need to evaluate the statistics of a neighborhood
of some selected pixels, then the Calculators are a
better approach.


If you are writing a Filter that needs to compute
statistics, you may use the Neighbor classes, ...
but this is more of an Advanced option...


In order to evaluate the trade-offs, you probably should
estimate first, in how many pixels of the image you need
to compute the statistics of their neighborhood. Is this
for 10% of the pixels in the image ? is it for 1% of them ?


   Regards,


       Luis


-----------------
IsabelleNg wrote:
> Hi ITK-users,
> 
> What's the proper way of calculating statistics in a neighborhood of X
> voxels in N-d? (i.e. standard deviation, mean intensity values in a
> neighborhood) Would these be the needed classes: MeanCalcuator,
> ImageToAdaptor, ListSample, etc.? Or, are there more efficient ways than
> instantiating sample per neighborhood visit? (i.e. use of vnl commands?)
> 
> Thanks,
> Isabelle