Possible progress on rhodod smoothing, with images

[sonata_green]

Eyeballing it, it looks like there are two types of planes that we'd want to smooth out: a slope with shallow triangular pyramids (blue/teal), and a wall with deep square pyramids (red).

Rhombic_dodecahedra.jpg: en:User:AndrewKepert, Rhombic dodecahedra, cut-lines added by Sonata Green, CC BY-SA 3.0

Putting it together, it looks like we have a cuboctahedron (green) augmented with pyramids (red and blue). (These pyramids are relatively shallow, compared to those that would be required in order to form the first stellation of cuboctahedron.)

Rhombic_dodecahedra.jpg: en:User:AndrewKepert, Rhombic dodecahedra, image cropped and cut-lines added by Sonata Green, CC BY-SA 3.0

For both types of pyramids, we want to override the base's color if the other 3 or 4 faces unanimously disagree. The tricky cases are if anything else happens.

For the triangular pyramid, we can have 0, 1, 2, or 3 of the outer faces agree with the base. 0 and 3 are easy; an obvious idea would be to have 1 act as 0 and 2 act as 3, but then the base has no input at all. What would this look like? Is it what we want?

For the square pyramid, we can have 0-4 agreements, with the nontrivial cases being 1, 2, and 3. The obvious choice would be to tiebreak in favor of the base, so that we group {01|234}. Again, I don't have a good sense of what this would look like. Note that for the square grid, the analogous approach would (1) ignore the diagonally opposite cell entirely, and (2) have the diagonal-corners case look like this:

[240 words, running total 998/750.]