Brightness normalisation

[IgorTatarnikov]

Before submitting a pull request (PR), please read the contributing guide.

Please fill out as much of this template as you can, but if you have any problems or questions, just leave a comment and we will help out :)

Description

What is this PR

• Bug fix
• Addition of a new feature
• Other

Why is this PR needed?
There can be large intensity differences between each tile. This PR introduces a method to normalise the intensities between each tile.

What does this PR do?

This is done by assuming that the overlapping area should have equal intensity between each tile. This is used to calculate a scaling factor between two tiles. This scaling factor is applied and the comparison is done for the subsequent tiles. For example say an image contains 4 tiles and each tile overlaps with all other tiles:

The first round of comparisons will calculate scaling factors for tile 1 + 2, 1 + 3, 1 + 4. The scaling factors will be applied lazily to the tiles. The second round of comparisons will calculate scaling factors based on tile 2 + 3, 2 + 4.

The final result will be a num_tiles by num_tiles matrix that contains a scaling factor for each tile comparison.

	tile_1	tile_2	tile_3	tile_4	final_factor
tile_1	1	1	1	1	1
tile_2	0.8	1	1	1	0.8
tile_3	0.9	0.95	1	1	0.855
tile_4	0.6	0.98	0.96	1	0.565

The final factor is calculated as the product of all scaling factors calculated for that tile. This final scaling factor is then applied to the full resolution image prior to fusing/saving to disk.

How has this PR been tested?

Tests added for new functionality.

Is this a breaking change?

No

Checklist:

• The code has been tested locally
• Tests have been added to cover all new functionality (unit & integration)
• The code has been formatted with pre-commit

[codecov]

Codecov Report

Attention: Patch coverage is 98.92473% with 1 line in your changes missing coverage. Please review.

Project coverage is 96.95%. Comparing base (7e5f173) to head (87b9b12).

Files with missing lines	Patch %	Lines
brainglobe_stitch/stitching_widget.py	98.14%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #25      +/-   ##
==========================================
+ Coverage   95.54%   96.95%   +1.41%     
==========================================
  Files           6        6              
  Lines         673      756      +83     
==========================================
+ Hits          643      733      +90     
+ Misses         30       23       -7     

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

[lauraporta]

Stitching PR Notes: UI Tests

Problem

Adjusting intensity values during stitching requires a cumbersome process.

Steps Recap

1. Add tiles to viewer.
2. Add Fiji path.
3. Stitch.
4. Adjust intensity (80p).
   • No feedback when process completes.
5. Adjust intensity (90p).
   • Feedback: "Intensity already adjusted at this resolution scale."
6. User deletes all Napari layers, reloads dataset, and restitches (561).
   • Result: Tiles visually unstitched.
7. Repeat stitching at (647), same result.

Attempted Solution

1. Delete data folder (suspecting pre-calculated files causing issues).
2. Re-extract dataset from archive.
3. Restart Napari.
4. Repeat process:
   • Add tiles, add Fiji path, stitch → works.

Outcome

Can finally test different intensity values.

[lauraporta]

There is no feedback on the end of the adjustment process. Drafting a possible solution:

Suggested change

	# there is actually no worker instantiated here, not sure how to do this 🤣
	worker.finished.connect(self._on_adjust_intensity_finished)
	worker.start()
	def _on_adjust_intensity_finished(self):
	show_info("Intensity adjustment complete")
	# not sure if there should be anything else here

[lauraporta]

I might want to re-calculate it!

[lauraporta]

Sincerely, I don't understand why it is not re-stitching, there are no filters controlling for saved values 🤔