Exporting RGB label image from VisuAlign, mapping to rainbow.json files fail

[Sverreg]

From Quicknii I can export the rainbow label image and get the area information from the RGB values. However, when I export the rainbow label image from VisuAlign, the RGB values (at least many of them) do not match with anything in the rainbow2015 or 2016.json lookup tables. Can someone point me in the right direction? I want to get the brain area for a given cell in my sample images that are aligned to the atlas RGB images, using the RGB values. Works for quicknii output, not for visualign.

Visualign output with non-matching RGB value highlighted:

[Tevemadar]

Those randomized colors are truly random, they're not meant for anything else than providing something to look at.
A thing you can try using is the atlas-recoloring "tool" at https://www.nesys.uio.no/VisuAlign/relabel.html - it produces an ITK label file with distinct colors. Depending on your workflow it may or may not be enough already. Let me know if you need the palette in a different format (like JSON).
Also depending on your workflow and personal preferences, you may or may not be interested in a Python code about working with the exported"flat" files, it's provided at https://www.nitrc.org/plugins/mwiki/index.php?title=visualign:Flat_file_format

[Sverreg]

Ok, thanks for the tip. But why does this behaviour differ between quicknii and visualign? Any chance it could be changed in visualign, to align the colors with the actual atlas?

[Tevemadar]

Ok, thanks for the tip. But why does this behaviour differ between quicknii and visualign?

The short answer is the format of the bundled atlases: QuickNII has its own binary format, VisuAlign uses NIfTI files coupled with ITK label descriptors, that's standard and can be used with other software.

Any chance it could be changed in visualign, to align the colors with the actual atlas?

That's what the images without the _rbw ending do: they have the colors from the original atlas, and that makes them unusable in case of mouse atlases from Allen: a single color is used for dozens of structures. It's somewhat visible in VisuAlign already, but if you're curious, have a look at the atlas and its structure tree in MeshView. The "rainbow" thing was created as a visual workaround.
What can happen in a future release is exporting the altered palette with the atlas slices, similar to QuickNII.

[Sverreg]

A thing you can try using is the atlas-recoloring "tool" at https://www.nesys.uio.no/VisuAlign/relabel.html - it produces an ITK label file with distinct colors.

So distinct colors, but not aligned to the atlas .json. If I understood it correctly. This does not solve the issue for me but:

Also depending on your workflow and personal preferences, you may or may not be interested in a Python code about working with the exported"flat" files

I always thought the flat files were just some kind of temp files used by the software, but this is exactly what I needed. The attached notebook processes a folder of .flat files and generates color-coded by atlas .png files. It looks correct for the handful of areas I sampled, and the only areas that appear to duplicate are opposite hemisphere areas, as expected. I'll attach it here in case anyone else has need of it.

Please let me know if you see any potential issues with this approach - otherwise feel free to close the issue.

FlatToPng.ipynb