CERN / TIGRE

TIGRE: Tomographic Iterative GPU-based Reconstruction Toolbox
BSD 3-Clause "New" or "Revised" License
529 stars 180 forks source link

How to do a FDK reconstruction without circle streaks? #481

Closed Lxy98Code closed 7 months ago

Lxy98Code commented 10 months ago

Hi! We're currently engaged in the process of FDK reconstructing a 3D object. However, we're encountering difficulties in eliminating the circular streaks around in the resulting reconstruction. Despite some efforts, we haven't yet identified an effective solution. If there are adjustable parameters that could potentially address this issue, we kindly request your guidance on how to proceed. Alternatively, is it possible to support FDK reconstruction without circle streaks?

Hope it make sense, Thanks!

AnderBiguri commented 10 months ago

Hi @Lxy98Code, can you show an image of the artifacts, so I can know what you mean?

The main reason TIGRE exists, is because FDK is a fairly bad algorithm, and unless working with perfect acquisition data, it tends to produce severe artifacts. This is why TIGRE has more that 10 algorithms, that should perform better. For example, you could try OS-SART.

That said, sometimes the cause of the artifacts can be some geometric mistake, so if you show me how they look like, I may be able to infer where they come from.

AnderBiguri commented 7 months ago

Closing it due to inactivity, feel free to open it again if you have issues