search

CERN / TIGRE

TIGRE: Tomographic Iterative GPU-based Reconstruction Toolbox
BSD 3-Clause "New" or "Revised" License
573 stars 188 forks source link

ComputeV fails for detector offset #301

Closed AnderBiguri closed 3 years ago

AnderBiguri commented 3 years ago

the 0.9 multiplier in computeV creates ring artifacts in detector offset recons.

Why is that there? I forgot.

yliu88au commented 3 years ago

I think it is there for keeping the back-projection of unity matrix (Vs) does not have zeros. Offset may make this logic fail, hence the artefacts. Should check if geox has correct settings.

AnderBiguri commented 3 years ago

@yliu88au thanks for the clarification!

I see and that makes sense. I think the better solution is perhaps to increase the size of the detector, instead of shirking the image size. I think that should be enough to make sure there are no zeros, while maintaining the correct pixel value in the V matrix.

AnderBiguri commented 3 years ago

nah, that does work to remove zeroes, but does not work with offsets.

AnderBiguri commented 3 years ago

Seems that the easy solution was to simply +0.000001 to V with standard geo.