New function: skimage.filters.try_all_threshold_dict (counterpart to try_all_threshold)

[GenevieveBuckley]

Description

A new function skimage.filters.try_all_threshold_dict is implemented.

An ordered dictionary is returned from skimage.filters.try_all_threshold_dict (keys = threshold method names, values = numeric value or None), whereas its counterpart skimage.filters.try_all_threshold returns a matplotlib visualisation.

>>> from skimage import data
>>> from skimage.filters import try_all_threshold_dict
>>> img = data.page()
>>> threshold_results = try_all_threshold_dict(img)
>>> for method, result in threshold_results.items():
...    print("{} threshold: {}".format(method, result))

Isodata threshold: 157
Li threshold: 148.1795556988551
Mean threshold: 171.54482984293193
Minimum threshold: 191
Otsu threshold: 157
Triangle threshold: 206
Yen threshold: 121

This is implemented as a separate function to skimage.filters.try_all_threshold for the reasons @jni outlines here:

As we are aiming to remove our matplotlib dependency in the medium term, probably the right approach is to have separate functions for visualising, rather than a keyword argument. This will make it easier to contain all mpl functionality in a single sub-package.

Checklist

[It's fine to submit PRs which are a work in progress! But before they are merged, all PRs should provide:]

• Clean style in the spirit of PEP8
• Docstrings for all functions
• Gallery example in ./doc/examples (new features only)
• Benchmark in ./benchmarks, if your changes aren't covered by an
  existing benchmark
• Unit tests

[For detailed information on these and other aspects see scikit-image contribution guidelines]

References

First suggested by @stefanv in 2178, and requested by @emmanuelle in 3149.

For reviewers

(Don't remove the checklist below.)

• Check that the PR title is short, concise, and will make sense 1 year
  later.
• Check that new functions are imported in corresponding __init__.py.
• Check that new features, API changes, and deprecations are mentioned in
  doc/release/release_dev.rst.
• Consider backporting the PR with @meeseeksdev backport to v0.14.x

[pep8speaks]

Hello @GenevieveBuckley! Thanks for submitting the PR.

• In the file doc/examples/segmentation/plot_thresholding.py, following are the PEP8 issues :

Line 65:1: E402 module level import not at top of file
Line 78:1: E402 module level import not at top of file
Line 79:1: E402 module level import not at top of file

• In the file doc/examples/xx_applications/plot_thresholding.py, following are the PEP8 issues :

Line 46:1: E402 module level import not at top of file
Line 47:1: E402 module level import not at top of file

[sciunto]

I don't find this name particularly elegant. Ideally, I would prefer try_all_threshold -> dict and try_all_threshold_plotted -> mpl. But this would require a deprecation cycle...

[stefanv]

Agreed; have try_all_threshold always return results, and then add a visualize flag?

[jni]

@stefanv that makes it harder/impossible to return the figure and axes as done currently. At any rate, if long term we want to remove the mpl dependency for analysis, we could have try_all_threshold return a dict, and then have a separate function in e.g. the viewer (or whatever we call the skimage-mpl package) to visualise that dictionary.

Either way, yes, a deprecation is going to be needed. =\

[sciunto]

It would make sense to me to do not duplicate this list in the code in order to make the addition of algorithms more straightfoward.

[sciunto]

I made 2 comments, but it's open for discussion.

[stefanv]

I like that suggestion--calculation wrapped by plotting in a separate function.

[alexdesiqueira]

Hey all,
I'm arriving kinda late to the party, but here's my suggestion.
Why not joining the three approaches: try_all_threshold() returning a dict (as @GenevieveBuckley and @stefanv proposed), and put an appropriate function to plot all of it in our viewer, such as viewer.view_all_thresholds() (@jni's suggestion)? This second function could have more structure, receive nrows, ncols, and so on.
We'd have two functions easier to handle, and our documentation could approach everything in a nice way.