[Nexus] Telco
"V. Armando Solé"
sole at esrf.fr
Wed Nov 7 14:46:35 GMT 2012
On 07/11/2012 15:36, Wintersberger, Eugen wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> Hello
> thanks Armando for your fast response.
>
> On 11/07/2012 03:14 PM, "V. Armando Solé" wrote:
>> Hello,
>>
>> I was at another (local) meeting.
>>
>> From the solutions summarized by Mark (good job there), we are
>> clearly in favor of the general case: there are additional data
>> containing the actual positions of the different pixels. Data
>> analysis software has in any case to use those positions, so why
>> not to get them directly from the source. That approach also allows
>> to deal with the case the modules form a non-flat arrangement
>> (adding an additional z component). About the fact that "destroys"
>> the image, one could argue the image was never existing in the sens
>> of a flat set of regularly spaced pixels.
> This is an interesting approach to the situation and I would like to
> add some more words to this. On the first glimpse I would prefer the
> suggestion of Tobias. However, when it comes to 'easy plotting' we may
> should think about what 'easy' means. I am actually working on a
> detector viewer application (it is a prototype for a GUI program to
> visualize data from Nexus files) using VTK. Indeed, when using VTK the
> general approach is much easier since VTK is clever enough to arrange
> the different module images to a large image by itself no matter how
> they are oriented in space (even a 3D view would be possible in the
> case of a curved detector).
>
> Thus, the question is rather: what means easy plotting for us?
In my case, if I have the coordinates, it is as simple to visualize
using OpenGL as visualizing images. If I have to visualize a curved
structure, then to have the coordinates is the simplest way: I can plot
the pixels on their own or I can make a Delaunay triangulation to
visualize them as a surface.
I guess now you understand why two years ago I insisted on having
interpretation="vertex" as one of the possible values for the
interpretation attribute of a dataset...
I would like to add that if the goal is to easily visualize the data,
one can even visualize them without correcting their positions. That
does not prevent users from setting rois on the different modules,
inspecting pixels and so on. Data analysis is other business, and there
you really need the position of the pixels. If one can get them from the
detector, why should one bother implementing different approaches?
Armando
More information about the NeXus
mailing list