search

odlgroup / odl

Operator Discretization Library https://odlgroup.github.io/odl/
Mozilla Public License 2.0
371 stars 105 forks source link

Class or helper functions to specify detector #353

Closed ozanoktem closed 8 years ago

ozanoktem commented 8 years ago

I'm not sure if this is part of ODL or not, but I still write it as a missing feature. I think we need a class or helper functions for specifying detectors. The geometries allows one to specify the detector surface (flat, circular, cylindric, ...) and the pixelization of that surface is given by a discretization of the coordinates for the points on the detector surface. Now, detectors may contain dead areas as in many emission tomography set-ups. These are merely regions between detector blocks. One should have a higher level class where the detector layout is specified and this will generate the necessary discretization of the coordinates for the points on the detector surface.

kohr-h commented 8 years ago

We have Detector and subclasses. I think we can represent these more complex hierarchical detectors.

adler-j commented 8 years ago

So do we keep this issue open? I feel the original problem can be solved by right multiplication with a indicator vector, for more complciated cases the Detector class is where to go. I cant see anything in particular being done on this issue atm.

kohr-h commented 8 years ago

To me this is a non-issue.

ozanoktem commented 8 years ago

Judging from the comments, there is no need to introduce additional structure for defining response models associated with various detector set-ups. Is that right?

kohr-h commented 8 years ago

That could be easily added in concrete subclasses. Detector does not disallow that.

ozanoktem commented 8 years ago

Ok, then close this issue and we add things when needed. I expect the first use case will be SPECT and/or ET.