Here's a well done Chinese montage done with what looks like after effects. Question..how did they do the 3D section of the restaurant next to the river? Did they rotoscope every frame for that 3D effect at 1:46? Or is that an actual camera moving through the scene? Link below looks funny because it's chinese characters that couldn't translate. There's some very good ideas in this piece.
???????????10?1????????????? - YouTube (http://www.youtube.com/embed/1LPbUa-JhkE?rel=0)

Dave Hill did a similar thing a while back, and has blogged about how it was done. From what I remember, it sounded seriously complicated. You can see the result here:

Dave Hill Takes You INSIDE His Photographs 3D Style | Fstoppers (http://fstoppers.com/dave-hill-takes-you-inside-his-photographs-3d-style)

But if you have a hunt round, there is quite a bit of information about how it was done. My third world internet connection is playing go-slow at the moment, but I am sure you'll find it.

Hi Trevor, thanks for your reply! However, Dave Hill's technique is different from the Chinese version. Dave's photo layers are flat, and the camera moves through the layers to create the illusion of 3D. In the Chinese version, the perspective changes as the camera moves. Take a look at the table as the camera moves over it. It's a technique I've never seen before.

It doesn't change perfectly though. The people are still flat. You could probably build that from photos in After Effects or something, with planes working for the tables, or maybe some complicated vanishing point stuff (http://www.adobe.com/designcenter-archive/video_workshop/?id=vid0287).
Seems too laborious and inflexible for that though. I think that since the cuttings in the water are 3d, the whole scene is probably built in a 3d package of some sort, just with simple forms covered with photographic elements as textures and cutouts standing up for the people (the bowls might be 3d as well)

Thanks Murray, that was the most plausible explanation yet! My first guess was that they used flying camera video footage and corner pinned still photo images of the people sitting at the table.

In this video you can see a similar 3D affect starting at about 0:36. I'll ask the guy who put this together how he did it. I know it can't be that intricate of a process as he's not that an FX specialist.

Scary Cow Introduction Video on Vimeo

Hopefully back with some answers soon.

Thanks Garrett! After watching the video, I'm guessing he "faked" his 3D by telling the actors to freeze while he ran around them with his camera?

Warren, thanks for sharing. I ended up watching the whole thing. There is a lot of cool stuff in there :)