This repository contains a customized implementation of the Neural Network Filtered Backprojection (NN-FBP) algorithm devised by Daniel Pelt.

The original NN-FBP code is available at: (https://github.com/dmpelt/pynnfbp).

If you intend to use this software, please cite the original publication: Pelt, D., & Batenburg, K. (2013). "Fast tomographic reconstruction from limited data using artificial neural networks". Image Processing, IEEE Transactions on, 22(12), pp.5238-5251.

The NN-FBP implementation in this repository does not require the installation of the Astra Toolbox, which works with GPUs. The gridding backprojector is used instead.

Basic compilers like gcc and g++ are required. The simplest way to install all the code is to use Anaconda with python-2.7 and add the installation of the python package scipy, scikit-image and Cython.

Procedure:

1. Create the Anaconda environment (if not already existing): conda create -n pynn python=2.7 anaconda.

2. Install necessary packages (if not already installed): conda install -n iter-rec scipy scikit-image Cython.

3. Activate environment: 'source activate iter-rec'.

4. Download the repo: git clone [email protected]:arcaduf/pynngrid.git.

5. Go inside the folder and install the C code for the backprojector: python setup.py.

If setup.py runs without giving any error all subroutines in C have been installed and your python version meets all dependencies.

If you run python setup.py 1 (you can use any other character than 1), the all executables, temporary and build folders are deleted, the test data are placed in .zip files. In this way, the repository is restored to its original status, right after the download.

Go inside the folder "scripts/" and run: python run_test.py.