MIT invents way to get reflections from photos




Researchers at the Massachusetts Institute of Technology have developed an algorithm that in many cases the reflections can automatically filter out photos taken through a window. Often objects, such as the photographer, visible in pictures taken through a window.

The algorithm will be presented by the investigators in June at the Computer Vision and Pattern Recognition Conference to the public. The effect is based on the fact that photos are taken through a window with double glazing, view generally two virtually identical reflections which are just a bit apart. One of the researchers, Yichang Shih explains on the site of MIT that occurs in double glazing from a reflection of the image in both the inner and the outer window. The same effect occurs when windows of thick glass only.

The system only operates at double reflection, otherwise it is as yet almost insoluble because the algorithm makes a comparison between two different, but otherwise similar images in a window. Shih gives an example: “If A + B equals C, then how can you recover A and B from a single C that is mathematically challenging Then we simply do not have enough criteria to draw a conclusion..”

The second reflection is thus necessary, whereby the value of A must be the same as the value of B for a pixel which is at the same distance in a given direction. If this selection criterion is added, it’s a lot easier to A, B and C D loosen weeks.

However, there is not the complete solution from the comparison of pixels in this way. For this, the researchers took another research group from stable. This group assumes that an image photographed through a window a certain statistical regularity property, like so-called “natural” photos. The idea here is that the pixel level abrupt transitions in natural and built environments by humans are strange and that if a transition, they run along clear boundaries. That means that when a cluster pixels comprises a part of a blue and a red object, all on the one hand is bluish, and on the other side reddish. Still, that approach does not work very well.

There appeared during the investigation something to sit in the pixel approach, but the algorithm had previously do some things to learn. For this, the researchers used a technique developed at the Hebrew University of Jerusalem. By letting statistic loose on blocks of eight by eight pixels in 50,000 test plates could be calculated the correlation between the pixels and to be thus obtained a good result.

Shih hopes that if the algorithm is further improved, it eventually may have a place in normal photo software and, perhaps more importantly, to make robots better “see” in areas with lots of glass.

MIT reflection removal



In: A Technology & Gadgets Asked By: [19073 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »

Star Points Scale

Earn points for Asking and Answering Questions!

Grey Sta Levelr [1 - 25 Grey Star Level]
Green Star Level [26 - 50 Green Star Level]
Blue Star Level [51 - 500 Blue Star Level]
Orange Star Level [501 - 5000 Orange Star Level]
Red Star Level [5001 - 25000 Red Star Level]
Black Star Level [25001+ Black Star Level]