Encoding luminance is easy, but really need to encode boundaries to give a good effect. May need a dependency, or perhaps it'll be simple enough to do at low resolutions.