neu_proc.ops¶
The array-in, array-out half of neu-proc.
Every function takes an array and voxel parameters, and returns a new array. Nothing here
imports a store, dask or blockrun; a test checks that importing this package leaves
neu_vol, tensorstore and dask unimported.
Per-axis arguments are zyx.
Lengths are per axis, or a scalar applied to all three, since a physical length becomes different voxel counts per axis on an anisotropic volume:
sigma_vox=(1, 4, 4).
Use them directly, or through neu_lib.Piece.apply(), which refuses a shape change
without a new frame= and a dtype change without kind=:
piece.apply(dust, min_voxels=100)
piece.apply(opening, radius=2)