Parallel Richardson-Lucy Algorithm Implementation

[avalluvan]

This pull request contains a parallel implementation of the Richardson-Lucy algorithm. Although the scripts can be run out-of-the-box with the instructions given in the README file, it is not adapted to the COSIpy framework yet nor is it written to be imported with python notebooks. It is intended to run the script with appropriate changes to its dependent configuration file. I am initiating this pull request to invite code reviews and start a discussion on what is necessary to port this to COSIpy.

Link to the original file: https://github.com/avalluvan/Richardson-Lucy/

[ckarwin]

Thanks for making a PR, @avalluvan! For now, can you please change the PR to a draft, since it's not ready to be reviewed for merging yet? It sounds like @hirokiyoneda would be the best person to look over this.

[hiyoneda]

Thanks @avalluvan. I am checking the codes now and agree to make this PR a draft. Then, members can test and check how to use the parallel calculation. @avalluvan, if you can, please consider the following things:

• Can you make your code more general? The code is for Ti-44, but I would like to know if it can be applied to other datasets, especially the Crab data. In a different dataset, parameters should be set again like NUMROWS, NUMCOLS. It would be great if these things are done automatically.
• Can you consider about using some codes already implemented in COSIpy? For example, the output (reconstructed image) can be history.Histogram (or image_deconvolution.model_map.ModelMap)? The parameters for RL methods can be referred to the yaml file used in the example notebooks.

The parallel calculation parts are R M and R^{T} (D/E), which are related to how to handle the image response matrix rather than the RL algorithm itself. So, I want to reflect Anaya’s code on the response class (and other classes can also use it), though currently, we don't have the image response class.