Small deviations from highly organized geometries

[kohr-h]

In contrast to fully free geometries as in #1023 we should also think about a possibility to allow for small deviations from the ideal geometry, mainly for the purpose of alignment either after or during reconstruction. The idea here is to keep the structure in data space as implied by the ideal geometry, but providing some machinery to handle small deviations.

This has, of course, some complications, like ordering of translations and rotations. Not sure how to handle that. Any good ideas here?

I'm mostly thinking about

• rotation axis (global deviation, one vector)
• angle per projection
• source position per angle (like flying focal spot)
• detector position per angle (small vibrations and whatnot)
• detector orientation per angle

The other, perhaps simpler, possibility is to use the vector geometries instead and use the implied data space weighting from the ideal case. But that would screw anything that is defined only for the true ideal geometries, like fbp_op.

[adler-j]

See #420 for some work in this direction that has sadly stalled.

[JevgenijaAksjonova]

Hi! I am going to implement Flying Focal Spot! Does anyone have ideas on how to do it better?