Space filling curves might be able to give us 1 point per pixel if done correctly for squares, circles, and triangles. It won't work for complex post transformations because we would need to reason about derivatives to ensure that the transformed object also has one point per pixel, but if we can get rid of the chaos game, I would be happy enough for now.