I'd have thought google, would have used a combination of the depth maps produced from the laser range finders they use when they take the street view pictures. Then using the depth map offset the image as appropriate.
Even then it doesn't seem very effective, the navigation arrows seemed to stand out more than the actual street.
Considering that when you rotate the view, the image offset doesn't change at all, I don't think this represents any additional information beyond the standard 2D streetview.
Even then it doesn't seem very effective, the navigation arrows seemed to stand out more than the actual street.