What if you could isolate & extract moving objects from video in nearly real time?
My Google colleague Yael Pritch worked with the Disney Research team that has now demoed how to remove noise & blur, increase resolution, make things disappear, and even clone moving objects. Check it out (and stick with it):
Fast Company provides an overview of the project, and you can read the paper here.
One thought on “Amazing new video segmentation research from Disney”
Really nice work. Between that and the video from yesterday, it was making me think that, at least for non-action shots, in the “future”, instead of trying to hold really still to take a nice crisp picture, we may be just taking handheld video and then post-processing the crispness (and maybe DOF) into it.