New module idea: descreen/dehalftone/inverse halftone

[jywarren]

Halftoned images, esp old images, can be "descreened" to produce clearer versions than just blurring the halftone.

Some use neural networks and some use FFT math. Here's a nice image of descreening from this excellent page and paper (click for higher res comparison):

Closeup:

Research:

• https://dercuano.github.io/notes/processing-halftoning.html
• https://www.cse.cuhk.edu.hk/~leojia/projects/rollguidance/results_dehalftone.html
• https://github.com/ccc870206/InverseHalftoneColorization
• https://github.com/MenghanXia/InverseHalftoning \
• GIMP descreen plugin:
  • http://web.archive.org/web/20161118075437/http://registry.gimp.org/node/24411
  • http://www.albumartexchange.com/forums/viewtopic.php?f=2&t=1986
• Opencv: https://github.com/6o6o/fft-descreen

Maybe the FFT-descreen OpenCV library looks simplest and most promising?

https://github.com/6o6o/fft-descreen/blob/38777f8bef9ab4ff695cc6c2fa957d34ab70ee44/descreen.py#L1-L54

Could we refactor that and model on our other OpenCV modules, like our Blob Analysis module?

https://github.com/publiclab/image-sequencer/blob/main/src/modules/BlobAnalysis/BlobAnalysis.js