Google & MIT's New Algorithm Uses Edge Detection To Remove Reflections From Photos

A Computational Approach for Obstruction-Free Photography

A team of researchers from MIT (Tianfan Xue, Michael Rubinstein, Ce Liu and William T. Freeman) are teaming up with Google with to present a new algorithm that is able to extract photographic inconveniences such as glares and reflections from photographs. The algorithm can then reproduce the image free of any reflections, in addition to being able to create an additional image of the reflection itself. This kind of problem solving would be especially useful when shooting behind glass or a fence, for example.

The algorithm requires the photographer to take a very brief video of what they are trying to photograph. The algorithm then compares the frames from the video clip to establish a reference photo using edge detection.

You can read the full details of the project by delving into the team’s research paper, A Computational Approach for Obstruction-Free Photography and by checking out the video below…

Sample Photo

[ via TechCrunch ]

Jon Ellis

or.. open the window…
- Tac A Coluccio
  
  Look at how high the vantage point in that photograph is… Windows on sky scrapers usually don’t open.
- Anthony Mehlhaff
  
  Hahaha!
- Jon Ellis
  
  /facepalm
Michael Turcotte

A whole new plot device for cops shows.
Rafael Lago

I wonder how long will this take to show up on customer software, like lightroom.
Mark Harrison

Simple algorithm change angle of camera so it is not a factor in the first place or turn off background lights. If that doesn’t solve it choose another photo op.
JOhn C

This is pretty cool. I would hope you wouldn’t have this much reflection, but if you had a little ‘spot’ you couldn’t eliminate when shooting it would be handy. I’m amazed at how much of the reflection it can recreate, I don’t have a use for the reflection(yet), it’s just interesting how well it recreates it.