Have you every attempted to take a photo or video through glass? What about shooting through a fence or a tree branch? These common obstructions can easily ruin our footage but this new software can completely clean the scenes with just a few reference frames.
This video accompanied a paper titled "A Computational Approach For Obstruction-Free Photography." The creators write:
We present a unified computational approach for taking photos through reflecting or occluding elements such as windows and fences. Rather than capturing a single image, we instruct the user to take a short image sequence while slightly moving the camera. Differences that often exist in the relative position of the background and the obstructing elements from the camera allow us to separate them based on their motions, and to recover the desired background scene as if the visual obstructions were not there. We show results on controlled experiments and many real and practical scenarios, including shooting through reflections, fences, and raindrop-covered windows.
I'm excited to see this technology built into our photos and video editing software in the near future. It's certainly not perfect but it it's an incredible start to something that will certainly be taken for granted one day soon.
that algorithm she described, thats exactly how I would have done it!
well it only work from a series of photos into a series of photos, and the foreground (reflection) and background have to be still. i think the practical application is quite limited.
Like all technologies, you have to start somewhere.
what you mean? it's impossible to do this from a single photo, or with moving subjects.
I'm sure the idea of content-aware was at one time considered impossible from a single image
MMA Photographers rejoice!!!!! :D
Hah that's a great point
It starting to look like in CSI series ;-)
Great
A chunk of this research came out of Microsoft. Some interesting stuff has been coming out of Microsoft relative to still and motion work. Their time compression algorithms are quite impressive as well.