Soft morphological filtering of solar images

[nabobalis]

The idea comes from this paper.

Base:

• [ ] Replicating the training dataset. (Add noise manually to a set of images)
• [ ] Working and well implemented algorithm.
• [ ] Compared to known output (Have to find source code.)

Extra:

• [ ] Optimized as much as possible for memory and CPU time.
• [ ] 100% test coverage
• [ ] Documentation and a worked example.

[dokutagero]

@nabobalis I'm interested in this issue. I have a question though. In the original paper, in section 4.1 they describe how they prepared the data for training the genetic algorithm. Basically they add noise manually to a set of images and as far as I know they haven't shared the training/testing images, so what would be a good way to train and test in order to compare results?

[nabobalis]

Would probably have to replicate their training dataset as close as possible.

[dokutagero]

Perfect, I'll take a look at it then :)

[nabobalis]

#35 adds an example that uses https://github.com/astropy/astroscrappy and it shows promise with some parameter fidding. We could either close this in favour or still implement the original feature.