Meh doesn't need to work like that at all. Doubling to get to the close and then the final % would probably work just as well or better and I still doubt with modern scaling algorithms that it would actually matter. Mathematically doubling or even 4x scaling makes a lot more since since you aren't having to do any sub pixel interpolation. I would still bet if you just scaled and did any of these other methods and then did a difference the difference is so minute. I was going to put a feature in xRes when I was working at macromedia that did all types of different math types and it just wasn't worth the very minimal difference.
