A cynic writes...
I don't think manually replicating what seam carving does is very difficult, or that this is a major step forward. Photoshop's Heal tool is capable of determining similarities on a small scale. Stitching software does this on a bigger scale with a specific purpose in mind. It's then not a massive leap to find anomalies, e.g. a person, in a scene and decide that they're of interest - and chop out stuff that isn't.
The huge display from TED is pretty fluid, but I suppose all it requires is an image format that's capable of a sort of progressive display, kind of like how computer games render far away models in low detail then do better representations as necessary. That's just a sensible approach to computation, though nicely done. I suspect there's more to it than that, but not much - yes, there's huge amounts of data but never in detail on display at once, so you tackle it in a kind of tree-like fashion.
I like the Notre Dame thing because it bodes well for improving stitching software and the like. The best thing I see in that is the ability to take very different photos, i.e. weather/lighting/shadows etc, and still make something of it. Otherwise you could do a lot of it now with consumer software.
