One thing it can be used for is to easily see what kind of palette a painter is using.

Tons of artistic applications. I can’t think of anything this visualization gives that standard scopes in software like Nuke don’t provide, but those are not as intuitive or quick to read.

Color in image reproduction is notoriously difficult to measure by simply looking at a whole image. Much of what we perceive as color characteristics are contextually dependent. And while the characteristics may not vary, the level of contextual dependence seems to. Females are more likely to have more reliable contextually affected color perception than males.

This kind of visualization is useful to understand why filming an actor in front of a green screen makes sense for special effects.

If you look at such an image, you will have two clusters of pixels, one for the green screen, and one for the actor. By selecting the appropriate cluster you now know which pixels you need to replace for compositing your actor in front of a CGI background.