Add resampling utility

[clbarnes]

Take a list of slabs.

Either resample every slab at a given length (see pynblast), linear/gaussian/spline, or distribute n sampling points proportional to length of each slab (randomly with seed?). Do not bind sampling points to ends, e.g. numpy.linspace(0, 1, n_samples + 2)[1:-1]

Interpolation will have to be implemented.

[clbarnes]

Possible algorithms

Resample with length

Arguments: criterion length

• Dendrites over-weighted if every slab is guaranteed a sample
• multi-branch points under-weighted if not
• resample algorithm - linear? gaussian? bezier?
  • doesn't really matter, linear probably fastest
• fixed length means larger neurons are over-sampled

Slab-length weighted

Argument: number of points

• assign points to slabs proportional to their length
• possibly randomly? deterministic at least
• resample slabs with that number of points

Descend from root, laying down treenode every X distance. At branch point, distribute remaining treenodes proportional to cable distance below branch. Probably over-weights terminal dendrites.

Shape simplification

• requires tree topology
• fixed branch and termination points over-weights dendrites

Ramer-Douglas-Peucker

• epsilon is hard
• not possible with non-topological data

Visvalingham-Whyatt

• keep removing smallest-area triangle until completion
• not possible with non-topological data (i.e. not skeletons)

[clbarnes]

Take proximal branches

Arguments: number of branches to keep

• requires tree topology

N.B.

This is probably a better fit for arbor-rs.