MIT researchers have developed an algorithm that can automatically remove window reflections from digital images.
Photographers usually only have two choices when they want to take pictures through a window. Either they open the window and take the pictures unobstructed by the panes, or they have to deal with the panes and the reflections that go with it.
There are ‘hacks’ that can lessen the impact of reflections from glass panes, but they’re not at all feasible at all times. Duct taping your lenses to the window and then covering the whole set up with black cloths helps, but who wants to keep doing that every time they need a better angle?
To solve this dilemma, researchers have developed an algorithm that takes into account the fact that modern windows are usually equipped with double panes. This method of insulation provide duplicate reflections—one on the inside and one on the outside. The nearly identical reflections offset each other by just a bit, giving the algorithm just enough information to figure out whether or not images in a photograph are reflections.
Researchers are proposing much broader uses for this concept. For instance, the algorithm can be applied to robotic sensors to help them distinguish reflections from the real thing.