[Paraview] Cells and Points in a Structured Grid Volume

[Christian Werner]

Thank you very much. Now things have become a little clearer. Anyway 
there are a few things which I do not yet understand.

I see now from the source of the Threshold plugin that the data to be 
processed is actually obtained in three different ways:

vtkPointData *pd=input->GetPointData(), *outPD=output->GetPointData();
vtkCellData *cd=input->GetCellData(), *outCD=output->GetCellData();

vtkDataArray *inScalars = this->GetInputArrayToProcess(0,inputVector); 
//inputVector is the input information vector


where the latter seems to be the array containing the actual volume 
data. The decision if a cell or point goes through the thresholding 
criteria is made a little later:

keepCell = this->EvaluateComponents( inScalars, ptId );
or
keepCell = this->EvaluateComponents( inScalars, cellId );


This raises two questions:

1) are the ptId and cellId disjunct?
2) if yes, does that mean that the InputArray contains point AND cell 
data? This clearly would require additional memory which is bad if you 
are working with big volumes.

Also I see the use of a DataSet which seems to hold all the cells of the 
volume and each cell seems to contain points ( these must be 8, right? ) 
I guess these points store coordinates? At least there is still another 
array containing only IDs. It seems as if there actually is a bunch of 
redundant information stored when I read structured grid, or at least 
when executing the Threshold filter.

This would explain why the memory usage skyrockets when manipulating our 
700MB (small version!) volume.

I don't want to complain here about anything, that memory issue is 
another story. More important to me is that I understand the structure 
of this.


Best regards,
Christian




Moreland, Kenneth wrote:
> In a 3D structured grid, there is the concept of points and cells and 
> VTK does in fact differentiate the two. In particular, 8 points make 
> up a hexahedral-shaped voxel cell. As is dictated by the topology, 
> there is one fewer cell than points in each dimension. For example, if 
> the grid has 512x512x128 points, it will have 511x511x127 cells.
>
> The raw image reader simply assumes you want to load the data as point 
> data and creates the grid accordingly. I don’t think there is any way 
> to make it load the data as cells. I can think of three ways to handle 
> this.
>
>    1. Close your eyes, take a deep breath, and find the inner peace to
>       simply let ParaView/VTK treat the data as point data. I cannot
>       think of any operation on cell data that can not be done on
>       point data. You just have to realize that ParaView will also let
>       you do some operations that may not be appropriate for cell data
>       (particularly those that interpolate the values in the cells).
>    2. Make a filter that converts the structured grid that has point
>       data to one that has cell data. It would simply create a new
>       topology on the output with one more point in every dimension
>       and then shallow copy the data as cell data. It should not be
>       too hard to implement. No, I am not volunteering.
>    3. Make a new reader that reads the data as cell data as it should.
>
>
> I hope that helps.
>
> -Ken
>
>
> On 2/26/10 8:45 AM, "Christian Werner" 
> <christian.werner at rwth-aachen.de> wrote:
>
>     Hello!
>
>     I am studying the Threshold Filter to dive more into the mechanics of
>     filtering and I just want to make sure I got the meaning of points and
>     cells right, especially when I have a structured grid volume as input.
>
>     Such input comes with some typical raw volume file which consists of,
>     lets say, 512x512x128 voxels all of them having some value.
>
>     Is it that this value is actually stored in a vtkPoints list that
>     contains uniformly distributed points in the volume? And a vtkCell in
>     this case gets the same value as the point because it contains
>     only this
>     single point in its center? ( with value I mean what you get from
>     GetComponent(id,c) )
>
>     So if I had some arbitrary (non-structured) input, a cell's value
>     would
>     be the average of all points lying inside the cell and changing
>     the cell
>     layout, i.e. making all cells bigger would accordingly change the
>     cell-point relationship and thus the cells value?
>
>
>     Best regards,
>     Christian
>     _______________________________________________
>     Powered by www.kitware.com
>
>     Visit other Kitware open-source projects at
>     http://www.kitware.com/opensource/opensource.html
>
>     Please keep messages on-topic and check the ParaView Wiki at:
>     http://paraview.org/Wiki/ParaView
>
>     Follow this link to subscribe/unsubscribe:
>     http://www.paraview.org/mailman/listinfo/paraview
>
>
>
>
> **** Kenneth Moreland
> *** Sandia National Laboratories
> ***********
> *** *** *** email: kmorel at sandia.gov
> ** *** ** phone: (505) 844-8919
> *** web: http://www.cs.unm.edu/~kmorel <http://www.cs.unm.edu/%7Ekmorel>
>