I agree with you that it starts to get a little old; it is not a new topic but there is something else that is really impressive: the result. This paper shows that we are moving forward on a complex topic like colour transfer between images. Compliments for the research work!
True, but a lot of those earlier examples relied on hand-labelled segmented images ("sky", "background", "person" etc.) whereas this could be directly used as an artist's tool without relying on academic data sets.
I'm personally hoping for some open-source code that will allow me to take photographs and game-engine stills and transfer the colour grading to enhance realism.
If all you're interested in is looking at pretty pictures, I can see how you'd say that. But from a research perspective, this area won't be "old" for a long time to come.