This work is closely related to Disney's Deep Canvas, about which Disney
was very secretive at the time. The fact that "Deep Canvas" existed
had been mentioned in the press, and there was apparently
some paper they'd released about it, which disclosed relatively
few details. Even having collaborated with someone at Disney
a little earlier, I was unaware of the existence of the project.
Anyhow, perhaps it's best seen as work that took some
ideas similar to those in Deep Canvas and provided details
about how to implement them.
One of my favorite bits of this work is in the stroke synthesis.
It wasn't complex, but it was (I think) the first place where
there appeared
a reasonable notion of describing a stroke in terms of neighborhoods
suitable for MRF/Machine-Learning-like techniques to be applied. Since then,
there's been some great work done, notably Pascal Barla's work
on pattern synthesis (in which he generates areas of ornamental
pattern from small examples).