I'm unclear on the novelty of this.  People have been dynamically overlaying video onto 3d models for a decade or more.  Indeed it is not too hard to interpolate between sufficiently close videos to change the vantage point (rather than the clumsy cross fade used in this).  And the idea of adding live video to google earth is something that my mother has thought of (we like to joke 'I can't see you on google earth dear, can you wave?').

3 or so years ago people were automatically generating the 3d montages directly from the photo data off flickr or whatever.

A nice video, but hardly ground breaking.