[Insight-users] NeighborhoodOperator

[l franz]

On 9/2/05, Joshua Cates <cates at sci.utah.edu> wrote:
> 
> Hi Franz,
> 
> There are, in fact, no such methods as GenerateScalarCoefficients() and
> ScalarFill(). The documentation is in error. You should implement the 
> methods
> GenerateCoefficients() and Fill() instead.
> 
> There is no requirement that the neighborhood operator be directional.
> GenerateCoefficients() takes no input and returns a vector of the 
> coefficients
> used in your operator. Fill() takes that vector of coefficients as the 
> argument
> and positions them spatially in the operator (which is, itself, a 
> neighborhood
> data structure).
> 
> For example, if you simply wanted to sum all of the values in a 
> neighborhood,
> your GenerateCoefficients() method would return the vector
> {1,1,1,1, ..., 1}. The Fill() method would simply copy those values into 
> the
> operator.
> 
> Hope this helps,
> 
> Josh.
> 
> l franz wrote:
> > Hi everybody,
> >
> > I want to create a neighborhood operator which inherits of
> > "NeighborhoodOperator" class. My operator will weight each pixels of a
> > NDimensional neighborhood through a kind of inner product (the output of
> > this inner product will be a neighborhood with weighted pixels). So my
> > operator performs on all the neighborhood (and not in a particular
> > direction) and its size is equal to the neighborhood size.
> >
> > So, my operator is not directional. In this case and if I understand
> > well the documentation, I have to implement two methods:
> > 1. GenerateScalarCoefficients
> > 2. ScalarFill
> >
> > So, I don't need to implement the "GenerateCoefficients" and "Fill"
> > methods (for directional operators) like in GaussianOperator or in
> > DerivativeOperator.
> >
> > But, I don't know how to create the "GenerateScalarCoefficients" and
> > "ScalarFill" methods. What do they have in input arguments? What do they
> > return? Do they have to call a particular method inherited from
> > NeighborhoodOperator??
> >
> > I think that "GenerateScalarCoefficient" method is similar to
> > "GenerateCoefficients" method. It takes no input argument and return a
> > "CoefficientVector" with the desired coefficients.
> > But I have no ideas of how to implement the "ScalarFill" method.
> >
> > Does someone know about it? Or does anybody know where I can find an
> > exemple of an non-directional neighborhood operator??
> >
> > Thanks,
> > Cheers,
> >
> > Franz
> >
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > Insight-users mailing list
> > Insight-users at itk.org
> > http://www.itk.org/mailman/listinfo/insight-users
> 
> --
> Josh Cates
> SCI Institute / School of Computing
> University of Utah
> 801-595-9165
> http:/www.sci.utah.edu/~cates



Thanks for your help Josh. That light me a lot.

But I still have a little question. I have implemented a 
"GenerateCoefficients()" and Fill()" methods. If I take in exemple the way a 
directional operator (Gaussian or Derivative Operator) is used, the 
coefficients are computed when the "CreateDirectional()" method is called. 
This last method is inherited form the "NeighborhoodOperator" class.

In fact, the "CreateDirectional()" method will call the 
"GenerateCoefficients()" and "Fill()" method that have been implemented. But 
this "CreateDirectional()" method needs the "m_Direction" variable (see 
NeighborhoodOperator.txx line 64). This last variable can be set by the 
"SetDirection(unsigned int)" method.

In my case, I don't have a directional operator. So it's not representative 
to set a direction.

So, have I the obligation to use the "CreateDirectional()" method to 
generate the coefficients? Or perhaps another method (perhaps 
"CreateToRadius" method).

And if I have to use the "CreateDirectional()" method, what would the 
direction to set? Can I simply "forget" to set the direction? (I think 
no...)

Thanks for your help,

Cheers,

Franz
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20050902/faf458b6/attachment.html