brainbow color clustering

python reimplementation for brainbow neural tracing:

[1] Duan et al., 2021, paper code

[2] Sumbul et al., 2026, paper

(1) denoising

• bm4d for each channel
• call separately as it needs long time and a lot of RAM

(2) supervoxelize, reimplement from [2]

(2.1) watershed on sobel

(2.2) add thresholded foreground

--> which color proximity function?

(2.3) split supervoxel with different colors

--> not implemented yet

(2.4) demix/merge supervoxel

--> not implemented yet

(3) GMM color cluster, from here [1]

• which color distance function?
• how to define dimension d for feature X^(Nxd)?
• how to define number of components for GMM?

(4) bridge skeletons

--> not implemented yet

• should be simple to implement to verify which cluster should be connected
• which distance threshold?

other todo's

• pipeline script
• gridsearch for hyperparameters on validation set
• submit evaluate instance segmentation