NSI to MBIR parameter conversion

[dyang37]

This PR contains the updates regarding the parameter calculation from NSI parameters to MBIR parameters.

With this update, we are able to consistently produce high-resolution MBIR reconstructions for NSI datasets.

Interface change

Function "NSI_load_scans_and_params" accepts a new variable "geom_report_path", which specifies the path to "Geometry Report.rtf" file.

Implementation

This PR contains a lot of changes, but the core changes are the following two lines in "preprocess.py": These two lines correspond to the following two functions that we discussed and agreed on:

Dependency change:

A new dependency "striprtf" is added to "requirements.txt". This package is used to read rtf file.

Image quality testing:

As we saw in my weekly slides, the image resolution is tested with five datasets. In all cases the nominal MBIR parameters produce the best reconstruction quality and resolution.

Demo scripts:

The NSI demo scripts are updated accordingly. you may run "demo_nsi_preprocess.py" to ensure that there are no bugs in the code. I sent out the updated datasets in a separate email.

[dyang37]

Close this PR for further modifications.

[dyang37]

Modification finished. PR is ready for testing. You may test the code by running demo_nsi_preprocess.py. This demo takes ~20 minutes on a Mac studio.

[dyang37]

Closing the PR for some final modifications.

[dyang37]

I modified the comments and the demo scripts based on the feedback this morning. Could you test the code by running "demo_nsi_proprocess.py", and accept the PR?

[cabouman]

I ran the demo, and it worked fine. I'm going to accept this PR. I'm sure it can be improved, but it's good enough, and we can make further improvements with additional PRs.