create a recipe for image processing

[toloudis]

Use Case

Currently there is no standard data flow to specify which channels of which 3d volumes will be recombined for presentation in CFE. In particular, we need a standard data structure to specify for preprocessing/recombination.

Also: do we want to allow for separated segmentations from raw data, or continue to preprocess for single images?

Solution

devise something like

combined_channels = [
 ["img_path", channel_index_or_name] , ...
]
crop_regions = [[x0,y0,z0, x1,y1,z1], ...]
region_names = ["cellid0", "cellid1",...]

to be used as input

[toloudis]

Q: is there a generic recipe format that will work and still allow optimal read and write speed or will the use case always depend on inverting loops and collecting up like things to be optimal