Ability to standardize aicsimageio output to 6D

[MosGeo]

Use Case

A standard predictable output from aicsimageio is needed for designing functions (and napari plugins) that will easily work on multiple types of image formats.

aicsimageio is great as great reader that produces a standard 5D array for all types of images. Currently, CZI reader produces 6D array for RGB images (edit, I am noting the RGB tiff files produces 6D too). This is great but it is not consistent with the 5D standard for all other types of images.

Solution

Add the ability (optional) to standardize the output of aicsimageio to 6D. One proposed implementation (napari-aicsimageio issue) is to have fill_all_optional_default_dims boolean to force aicsimageio to produce the 6D array.

Alternatives

In each processing workflow, the functions needs to check if the image is 5D or 6D.

EDIT: Reading more about the issue, i noticed that aicsimageio are following the zarr standard (t,c,x,y,z) which is great. Now, it is the CZI that is not following that as they distinguish between RGB color channels and 'other' channels. I don't know what is the best solution. Is there a way to figure out if the channels represent RGB in the (t,c,x,y,z) format? just thinking out loud.

[toloudis]

Arguably aicsimageio is biased toward scalar intensity storage rather than RGB. The Zeiss CZI file format is an important one to cover and we didn't want to start mixing rgb channels with "other" image channels (e.g. different laser wavelength captures) because it would confuse a lot of things. We opted to go to TCZYXS where S is samples and dims["S"] = 3 for rgb.

The expected way for users of aicsimageio to know if they got a RGB image or not, is to either (a) know this before loading the image somehow, or (b) actually check the number of dimensions or check for the presence of a "S" dim.

One thing we considered is having some kind of mode or flag in AICSImage such that it always will add the S dimension. So if you create myimg = AICSImage(..., has_samples=True) then all standard data accesses would be expected to be 6D instead of 5D.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.