Take image A. Work various transformations and filters on it to create image B. Run the start point (A) and end point (B) through the software. The software figures out the gestalt of the transformations … and can now apply them to image C to create image D.
Cool stuff, in and of itself.
Now imagine A is a photo, and B is a watercolor painting of the same subject …
(Also via Boing Boing)