CIL vs CVXPY

This repository contains several scipts that compare the solution of different imaging minimisation problems using the Core Imaging Library (CIL) and cvxpy. All the cvxpy scripts are initially implemented in Matlab for [1] and [2].

Imaging Problems

• Total Variation Denoising
• Total Generalised Variation Denoising
• Directional Total Variation Denoising
• Total Variation Tomography Reconstruction

Note

• For the cvxpy implementation, the Splitting Conic Solver (SCS) is used by default. Another option is to use the MOSEK solver but it requires a licence. Institutional Academic License is free.

Run the notebooks on Binder

In order to open and run the notebooks interactively in an executable environment, please click the Binder link above.

Run the notebooks locally

Alternatively, you can create a Conda environment using the environment.yml in the binder directory:

conda env create -f environment.yml

References

[1] Infimal Convolution Regularisation Functionals of BV and L^p Spaces. Part I: The finite p case. Burger, Martin, Papafitsoros, Konstantinos, Papoutsellis, Evangelos, and Schönlieb, Carola-Bibiane Journal of Mathematical Imaging and Vision 2016

[2] Infimal Convolution Regularisation Functionals of BV and L^p Spaces. The Case p=∞. Burger, Martin, Papafitsoros, Konstantinos, Papoutsellis, Evangelos, and Schönlieb, Carola-Bibiane In System Modeling and Optimization 2016