[vtkusers] Fixing vtkPolyDataToImageStencil

[David Gobbi]

Never mind my earlier comment about the data.  I clicked around
on the github link you sent and found it.  Thanks muchly.

On Mon, Oct 6, 2014 at 1:48 PM, David Gobbi <david.gobbi at gmail.com> wrote:
> Also, just in case you are using the patch that was attached to
> the bug report, that patch is no longer up to date.  The most
> recent code can always be found on the gerrit submission page:
> http://review.source.kitware.com/#/c/17021/
>
> On Mon, Oct 6, 2014 at 1:44 PM, David Gobbi <david.gobbi at gmail.com> wrote:
>> I suspect there are problems with the way that the tolerance is
>> applied during the rasterization part of the algorithm, and the larger
>> the tolerance, the more likely it is that problems will occur. For
>> best results you should use a tolerance of zero.
>>
>> The patch is still sitting in the queue because I haven't had time to
>> test it properly.  If it is at all possible for you to provide data
>> sets or test code, that would be a big help.
>>
>>  - David
>>
>> On Mon, Oct 6, 2014 at 12:05 PM, Ricardo A Corredor
>> <ra.corredor at gmail.com> wrote:
>>> Hi David,
>>>
>>> I just made a test using my data and your patch. It seems it is working
>>> well, but I still don't get how the tolerance should be selected. I explain
>>> myself with the following image and description:
>>> <http://vtk.1045678.n5.nabble.com/file/n5728996/PolyDataRasterization.jpeg>
>>>
>>> I'm using a very simple modified version of the VTK Cxx example,
>>> PolyDataToImageData (code and data are here: http://git.io/s49TIA). I used
>>> two polydata (the cells of both polydata are polylines, i.e. closed contours
>>> along Z axis, that could be seen in the image in white), one reasonably
>>> bigger than the other one (in the image you will see the body and a
>>> structure inside), but keeping same properties. When using the Tolerance
>>> (0.001) by default in the big polydata (the body), the 'raster' image is
>>> generated correctly. When using the same code in the small structure, the
>>> output image still has strange blank lines (left part of the image).
>>> However, if I decrease considerably the Tolerance (0.00001), the result is
>>> correct (right part of the image).
>>>
>>> Does it makes sense to you? Is that normal?
>>>
>>> Thanks!
>>>
>>> Regards,
>>>
>>> Ricardo
>>> RaC