Video Epitome (Sort of a Super Resolution algo)

[Sagekilla]

I don't know if anyone else has seen this (http://www.psi.toronto.edu/~vincent/videoepitome.html) yet, but I remember reading about it a while ago (before I started coming to doom9). It's what I believe is basically a Super Resolution algorithm, thought you guys might be interested in it..

It sounds like it could prove useful, but I'm willing to bet it's insanely slow to use. Any thoughts on feasibility of a function based off this idea?

[MfA]

For the example they used they learned the epitome from the original high resolution sequence ... the work to turn it into a general superresolution method (where you obviously don't have the original sequence) is not trivial and well beyond what the paper attempts.

Patch based techniques in general are interesting (in theory you can implement non local filtering a hell of a lot faster than with say NL-Means) but I'm not sure how useful extending it into the time domain is.

[Sagekilla]

Mmm.. still, it would be interesting if any concepts or new functions could be derived from this.

[Terka]

Looks interesting!