I was checking the ASCII representation of the binary bits 7-51 of the SHA 1 and 256 hashes of the hint image. Maybe the original image? Anyway, the other problem with my theory is 45 = 9 * 5, so it doesn't break up evenly into bytes, 7-bit ASCII, hex nibbles, or really anything useful.
I don't think that would work from a reverse engineering perspective... in order to construct the puzzle, she'd have to change the puzzle, then the hashes would be wrong.
I'm personally leaning towards that thing in the middle being the cheshire cat, but I'm also a big Alice in Wonderland geek, so that might be part of it.
It is surprisingly challenging to extract clearer image data out of that picture. As soon as you desaturate it everything goes away, if you bump up the contrast it goes crazy, etc. The letters are relatively simple.