0

I'm trying to implement an optimization algorithm in Python for solving a computerized tomography problem with TV regularization. I know I can use the function "radon" from scikit-image, but the point it that I also need the transpose (or adjoint operator) of the Radon transform as well.

Does anyone know if there is a way to compute the transpose matrix of the Radon Transform in scikit-image?

Nabuco
  • 1
  • This is a software question, not a mathematics question. You might get an answer but in general you should direct it to the right forum for an answer. – rschwieb Feb 22 '22 at 18:21
  • I see you've already posted elsewhere on the web and also here. I know you're probably eager for an answer but nobody likes excessive crossposting. – rschwieb Feb 22 '22 at 18:22
  • Searching [scikit-learn] at stackoverflow transpose gets over 200 hits... any of those happen to have the solution? searching with scikit-image gets fewer hits, but you might check that too. – rschwieb Feb 22 '22 at 18:25
  • Ok, sorry for the inconvenience and thanks for your help and time. As you can see I'm new here (and in the other forums you pointed)... I promisse I'll be more carefull next time here (if any). Bye! – Nabuco Feb 22 '22 at 23:26

1 Answers1

0

By reading the documentation, I don't think that Scikit Radon transform has this feature. I would therefore not use it on my own, since the adjoint is the most needed operation in scientific computing. Browsing through the web, I saw at least one library (pylops) offering this feature: https://pylops.readthedocs.io/en/latest/gallery/plot_radon.html

The other option is to reprogram it as a sparse matrix and use standard matrix-vector products...

P. Weiss
  • 1
  • You got exactly the point: I have a sort of primal-dual method, so the adjoint is really crucial. pylops looks really interesting, I will take a look at it. Many thanks for you kind reply! – Nabuco Mar 10 '22 at 14:53