I would still argue that a too easy, call it lazy, brainwallet is kind of security by obscurity. You hide your too easy, too bad, maybe lazy, easy to remember phrase behind a hash and hope it's not found by "crackers". People tried to cut corners to be able to remember some random looking secret when in fact it wasn't hidden well enough.
You can have a perfectly safe brainwallet, but not derived from publicly available data or publicly available words which probably only need some amount of mixing. We all know how this turned out for many brainwallets.
Humans are bad sources for entropy. By brainwallet's scheme, if I were to generate a key that is sufficiently long and has sufficient entropy, it would never get cracked. Note that brainflyer cracks the keys at a pretty fast speed but it is very far from exhausting the entire space of SHA256.
The thing with this scheme is somewhat similar. You could try to make some abstract digital painting with only 8 or 16 colors. Would that attract suspicion? It depends, you can't be sure. Decoding the colors' hex values, transforming them to decimal could raise suspicion again because you could notice that all colors have at their front two digits monotonically rising in a quite unusual way.
I'd say, it's not an easy puzzle to solve but not well enough hidden, too. For hidden in plain sight, I believe, it's a gamble. With this, I don't like to gamble.
Hiding anything without encryption isn't the best way of doing things anyways. Even a safe can be cracked given time, and most people usually add a layer of encryption ontop of their keys before hiding behind something.
The beauty about this is that there isn't a set way of recovering or encoding your keys. It is perfectly possible for you to choose a unique obfuscation technique that doesn't raise the slightest form of suspicion and people wouldn't bother figuring out how you've encoded it. After all, art is abstract and a bunch of colours mashed together wouldn't raise any suspicion.