From Ruven's talk at the conference, a transform of the color space could be "negative". Useful for getting the positive of a digitized film negative, or (for some reason) the negative of a positive image.
(There were other qualities suggested too, which I didn't capture)