search

JuliaImageRecon / Sinograms.jl

Julia library for working with sinograms / tomography / Radon transform
MIT License
15 stars 6 forks source link

Parker Weights: Cone Beam - Short Angle #43

Closed tknopp closed 1 year ago

tknopp commented 1 year ago

After successfully reconstructing the slices of a flat panel cone beam dataset using the fan beam geometry (SinoFanFlat) I next tried out the full 3D dataset using CtFanFlat. Again, things went rather smoothly and it required just minor modifications which I could easily extract from the documentation.

However, with the FDK reconstruction I now get the same artifacts that I got with the fan beam geometry when using no parker weights. Therefore the question: Can it be that the parker weights for short angle 3D scans are currently not used?

Below is a screenshot. Upper left is the Sinograms.jl reconstruction and upper right the reconstruction provided by the scanner.

Bildschirmfoto 2022-12-23 um 10 50 51
JeffFessler commented 1 year ago

Can it be that the parker weights for short angle 3D scans are currently not used?

Correct - I had not gotten around to implementing or testing that. Sorry! I will work on it pretty soon.

tknopp commented 1 year ago

Thanks, no hurry!

tknopp commented 1 year ago

Fixed by https://github.com/JuliaImageRecon/Sinograms.jl/pull/46, thanks!

JeffFessler commented 1 year ago

I hadn't closed it yet because #47 is a major overhaul that should be better yet, and I'm planning to wait to tag a new release when that is done. Glad to hear #46 worked for you. Previously I had built lots of it on getproperty and that turns out to be bad for type stability. I'm completely eliminating that now...

tknopp commented 1 year ago

ok, then lets re-open.

JeffFessler commented 1 year ago

Ok, now I have merged #47 and it passes all the tests and the docs look to build correctly, so I'll close this and tag 0.4. The only annoying thing is that codecov does not seem to track Threads.@threads properly: https://app.codecov.io/gh/JuliaImageRecon/Sinograms.jl/blob/main/src/fbp3/cbct-back.jl so I have to settle for 99% instead of 100% coverage and the scarlet x for failed CI ☹️ It is quite possible that this massive overhaul will have broken something so when you have a chance please try the new version on you short CBCT data and let me know of any issues.