[Paraview] Point Dataset Interpolator - EllipsoidalGaussianKernel

[M P]

hmm, reading the vtkEllipsoidalGaussianKernel documentation, it seems it is
not suitable for my purpose after all. Not possible to achieve constant
weight inside the sampling cylinder. Also the use of word "normal" in the
documentation confuses me, e.g.: "z is the distance of the current voxel
sample point along the local normal N"

W(x) = S * exp( -( Sharpness^2 * ((rxy/E)**2 + z**2)/R**2) )

- I would expect that it should be direction of the curve, not normal. When
I display Normals on my spline curve it is as expected perpendicular to the
curve and that I believe should not be the z coordinate of the kernel.

Will make some simple Programmable Filter for that.

Best regards,

Martin

On Tue, Aug 22, 2017 at 11:36 AM, M P <martin34148 at gmail.com> wrote:

> Hello all,
>
> I am trying to setup this filter to sample a point volume dataset along a
> defined spline curve. The ideal "kernel" for my purpose would be a short
> cylinder with some relatively big radius and constant weight in the whole
> volume of the cylinder (the axis of this cylinder should follow the
> direction vector of the spline). From all the kernels available for this
> filter the "EllipsoidalGaussianKernel" seems to be the most suitable. But I
> have a problem understanding what some of the parameters exactly mean.
>
> ------
> "Use Normals": Specify whether vector values should be used to affect the
> shape of the Gaussian distribution.
>
> ?: What Normals vector values are meant here? Is it the normal to the
> spline or some vector from dataset? And how is then the kernel oriented if
> this option is not used?
> ------
> "Use Scalars"
>
> ?: Similar question, what scalar is meant here and how is the weighting
> done? But I guess this is not important for my use case
> ------
> "Sharpness": Specify the sharpness (i.e., falloff) of the Gaussian. By
> default Sharpness=2. As the sharpness increases the effects of distant
> points are reduced.
>
> ?: I would like all the dataset points have the same weight - so I guess I
> need maximum sharpness (like 20) - but this parameter help comment indicate
> that it works the opposite way - higher value means lower weight of distant
> points
> ------
> "Eccentricity": Specify the eccentricity of the ellipsoidal Gaussian. A
> value=1.0 produces a spherical distribution. Values less than 1 produce a
> needle like distribution (in the direction of the normal); values greater
> than 1 produce a pancake like distribution (orthogonal to the normal.
>
> ?: Here I am confused by the word "normal", should I understand that it
> actually means direction vector?
> ------
>
> Is there some paper or book where this parameters are exactly explained,
> or could someone point me to a source code where this is implemented (and
> hopefully it would be possible to understand from there)?
>
> Best regards,
>
> Martin
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20170823/71c4670d/attachment.html>