Curious if it would work in 3D

Very interesting! I wonder what that would look like

Right now, roughly, the algorithm recursively divides the image by doing decimation (ie. picking every other pixel), and keeps the decimated pixels as a second image

Not sure how that algorithm would apply to a 3d data structure

Do you know how 3d objects/images are usually represented?

It would be cool to recursively decompose a 3d object into smaller versions of itself :)