Idea: don't collapse junction pixels

[jni]

I've referred people to the junction-handling in skan from the paper's supplementary material more times than I can remember:

https://peerj.com/articles/4312/#supp-1

It's a rather confusing part of the code, and though it is useful, it would be good to have alternatives that preserve each and every pixel in the original skeleton.

One idea: of all the pixels in a junction, only assign one to be a junction (how to do that is left as an exercise for the reader). The others get assigned to the incoming paths, and all links to non-junction pixels are removed (ie the degrees are artificially reduced).

[jni]

of all the pixels in a junction, only assign one to be a junction (how to do that is left as an exercise for the reader)

Just had a thought: replace the junction pixels subgraph (which will be something close to a clique) with its minimum spanning tree. This is very well defined and I think it will do the right thing in most if not all cases.

[jni]

@kevinyamauchi @marlene09 next week is pretty busy for me but perhaps we can work on this the week after that? I keep coming back to this idea (junction -> minimum spanning tree), and I've convinced myself it's the right way forward! I think it'll solve all our problems with pruning/editing skeletons!

[marlene09]

@jni, @kevinyamauchi, I can do the week after, let me know when

[kevinyamauchi]

Next week works for me! Tuesday (Sept 7) or Wednesday (Sept 8) morning (CEST) work best for me.

[marlene09]

The 7th works better for me :) talk soon.