feat: support multiscales

[jo-mueller]

Hi @m-albert ,

small QOL feature to enable working with multiscales in napari that are not xarray multiscales. This makes the napari-multiview stitcher compatible with ome-zarr 0.6-backed scenes. In anapri, you can then just drag&drop an entire scene along with the coordinates of each tile, hit fuse or register and you're done :)

multiview_stitcher_example.mp4

[m-albert]

Hey @jo-mueller, this is super cool 😍😎🚀

Will have a look at the code tomorrow!

[m-albert]

Perfect, so with the changes you made the stitching plugin gets multiscale xarray images from multiscale napari layers 🥳🥳

That will allow the plugin to use the right resolution layer from the napari layer for registration. For fusion, napari-stitcher is a bit behind multiview-stitcher and still fuses individual resolution layers.

If you don't mind I'll hijack your PR branch and add changes to complete full multiscales support.

[jo-mueller]

If you don't mind I'll hijack your PR branch and add changes to complete full multiscales support

Not at all :) Actually maybe you could also hijack the resolution level selector in the latest napari version? I.e., you could set the resolution on a per layer basis and then use that as an indicator on which level to use? 🤔

[m-albert]

I like the idea! The closer the control is to vanilla / built-in napari the better :)

How about the following: If the resolution is set to a given resolution for a given layer, that's used within the stitcher. If it's left on Auto, the stitcher chooses the resolution.

[m-albert]

Okay, I

• adapted the code to use new multiview-stitcher multiscale features
• added the possibility for the user to choose whether fusion is computed on the fly or pre-computed to OME-Zarr
• made registration use the scale chosen by the user in the napari interface as per your suggestion @jo-mueller :)

I also added a multiscale sample dataset which illustrates the multiscale behaviour :)

multiscale_digits.mp4

[jo-mueller]

Woah awesomeI! Love the numbering in the scale levels, very cool idea 💪

[codecov]

Codecov Report

❌ Patch coverage is 95.48193% with 15 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.75%. Comparing base (f9f950e) to head (7cf06b4).
⚠️ Report is 10 commits behind head on main.

⚠️ Current head 7cf06b4 differs from pull request most recent head 52a5358

Please upload reports for the commit 52a5358 to get more accurate results.

Files with missing lines	Patch %	Lines
src/napari_stitcher/_stitcher_widget.py	88.00%	12 Missing ⚠️
src/napari_stitcher/viewer_utils.py	92.30%	2 Missing ⚠️
src/napari_stitcher/_mosaic_widget.py	96.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #22      +/-   ##
==========================================
+ Coverage   86.24%   88.75%   +2.51%     
==========================================
  Files          13       14       +1     
  Lines         858     1290     +432     
==========================================
+ Hits          740     1145     +405     
- Misses        118      145      +27     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jo-mueller]

Following this, it would probably be really cool to have a writer in napari so one could then save the stitched output into ome-zarr 🤔

[m-albert]

Definitely! Maybe napari-ome-zarr would be a good place for this, or maybe even already has a writer?

Next to the stitched image it'd be cool to be able to save the refined transformation parameters (without writing image data) to individual multiscales, or an OME-Zarr v06 scene.

[jo-mueller]

Napari-ome-zarr would be the right place, but can't do it yet 👍

[m-albert]

I see, do you know why it is that it cannot do it yet? Is it simply that it's not implemented yet or are there blockers?

[jo-mueller]

Totally in scope! I think since the more recent API changes at ome-zarr-py, it should be relatively straightforward. Project for Thursday?

Edit: ome/napari-ome-zarr#163

[m-albert]

Sounds great :)

[m-albert]

Merging here! Thanks for kicking this off @jo-mueller :)