MIT algorithm aims to eradicate reflections from photos taken through windows
Researchers at the Massachusetts Institute of Technology claim to have developed a method for eliminating glass reflections in photos via digital processing. It is hoped that with further development the idea could see its way into cameras themselves, enabling reflections to be automatically removed when they interfere with the view through a window.
The new algorithm, due to be formally announced in June at the Boston Computer Vision and Pattern Recognition conference, is claimed to offer a way of distinguishing the image reflected by a window surface from the image of whatever is behind the window. Once the two images are clearly identified one can be subtracted from the whole to leave the other.
MIT researchers noted that objects reflected by glass surfaces tend to have two reflections, not one, and that the two images of reflected objects are slightly off-set. The multiple images in a reflection come about sometimes because windows are double glazed for insulation and thus have two panes of glass, and other times because light reflects from the rear surface of a pane of glass as well as the facing surface.
Further research showed that the interference of a reflection can often be identified by the way it interrupts the regular shape and color patterns of natural and man-made objects on the other side of the glass. Working with small clusters of 8x8 pixels, researchers analyzed color and tonal relationships within the group and came to a method by which certain data correlations indicated that the group included detail from a reflection.
The result is that when an image of, for example, a view shot through a bus window is analyzed, the team can separate the image of the view from the reflected image of the photographer that shot the view – and then can show either the view on its own, or the photographer on his or her own.
If successful the algorithm may have serious consequences for the flexible rubber lens hood market, though early indications suggest polarizing filter manufacturers will be safe for some time to come.
For more information see the Massachusetts Institute of Technology website.
Removing reflections from photos taken through windows
New algorithm exploits multiple reflections in individual images to distinguish reflection from transmission.
Larry Hardesty | MIT News Office
May 11, 2015
It's hard to take a photo through a window without picking up reflections of the objects behind you. To solve that problem, professional photographers sometimes wrap their camera lenses in dark cloths affixed to windows by tape or suction cups. But that's not a terribly attractive option for a traveler using a point-and-shoot camera to capture the view from a hotel room or a seat in a train.
At the Computer Vision and Pattern Recognition conference in June, MIT researchers will present a new algorithm that, in a broad range of cases, can automatically remove reflections from digital photos. The algorithm exploits the fact that photos taken through windows often feature two nearly identical reflections, slightly offset from each other.
“In Boston, the windows are usually double-paned windows for heat isolation during the winter,” says YiChang Shih, who completed his PhD in computer science at MIT this spring and is first author on the paper. “With that kind of window, there's one reflection coming from the inner pane and another reflection from the outer pane. But thick windows are usually enough to produce a double reflection, too. The inner side will give a reflection, and the outer side will give a reflection as well.”
Without the extra information provided by the duplicate reflection, the problem of reflection removal is virtually insoluble, Shih explains. “You have an image from outdoor and another image from indoor, and what you capture is the sum of these two pictures,” he says. “If A+B is equal to C, then how will you recover A and B from a single C? That's mathematically challenging. We just don't have enough constraints to reach a conclusion.”
Thinning the field
The second reflection imposes the required constraint. Now the problem becomes recovering A, B, and C from a single D. But the value of A for one pixel has to be the same as the value of B for a pixel a fixed distance away in a prescribed direction. That constraint drastically reduces the range of solutions that the algorithm has to consider.
Nonetheless, a host of solutions still remain. To home in on one of them, Shih and his coauthors — professors of computer science and engineering Frédo Durand and Bill Freeman, who were Shih's thesis advisors, and Dilip Krishnan, a former postdoc in Freeman's group who's now at Google Research — assume that both the reflected image and the image captured through the window have the statistical regularities of so-called natural images.
The basic intuition is that at the level of small clusters of pixels, in natural images — unaltered representations of natural and built environments — abrupt changes of color are rare. And when they do occur, they occur along clear boundaries. So if a small block of pixels happens to contain part of the edge between a blue object and a red object, everything on one side of the edge will be bluish, and everything on the other side will be reddish.
In computer vision, the standard way to try to capture this intuition is with the notion of image gradients, which characterize each block of pixels according to the chief direction of color change and the rate of change. But Shih and his colleagues found that this approach didn't work very well.
Playing the odds
Ultimately, they settled on a new technique co-developed by Daniel Zoran, a postdoc in Freeman's group. Zoran and Yair Weiss of the Hebrew University of Jerusalem created an algorithm that divides images into 8-by-8 blocks of pixels; for each block, it calculates the correlation between each pixel and each of the others. The aggregate statistics for all the 8-by-8 blocks in 50,000 training images proved a reliable way to distinguish reflections from images shot through glass.
In their paper, Shih and his colleagues report performing searches on Google and the Flickr photo database using terms like “window reflection photography problems.” After excluding results that weren't photos shot through glass, they had 197 images, 96 of which exhibited double reflections that were offset far enough for their algorithm to work.
“People have worked on methods for eliminating these reflections from photos, but there had been drawbacks in past approaches,” says Yoav Schechner, a professor of electrical engineering at Israel's Technion. “Some methods attempt using a single shot. This is very hard, so prior results had partial success, and there was no automated way of telling if the recovered scene is the one reflected by the window or the one behind the window. This work does a good job on several fronts.”
“The ideas here can progress into routine photography, if the algorithm is further robustified and becomes part of toolboxes used in digital photography,” he adds. “It may help robot vision in the presence of confusing glass reflection.”
For most of Managing Editor Allison Johnson's photography, smartphones have already replaced a traditional camera. But a recent trip reinforced a couple of key reasons why she's not ready to quite ready to leave the dedicated camera at home – yet.
In the final part of our beginners' guide to camera fundamentals, we look at the trade-offs you contend with when you choose a sensor size. We hope it helps you find the balance that works best for you.
The roof and spire of Notre Dame Cathedral were destroyed in a devastating fire this past Monday. Drone footage shows the extent of the damage done to the historic Parisian landmark.
Fujifilm's latest firmware update for its X-T3 camera includes an improved AF algorithm for enhanced face detection, better subject tracking and more.
Zhong Yi Optics has released its new Mitakon Speedmaster 50mm F0.95 III lens for Sony FE, Canon RF, and Nikon Z mount camera systems.
You can now see the artwork on your wall without the need to order it first.
Nikon's Z6 offers 24MP full-frame image quality in a tough and lightweight body. Photographer Diego Rizzo took a Z6 to Guatemala recently to shoot the Fuego volcano - watch our video to see how he got on.
The Fujifilm X-T30 is a $900 camera that's an incredibly capable stills/video hybrid. Image and 4K video quality are top-notch and, aside from some ergonomic and autofocus stumbles, the X-T30 does just about everything right. Learn more in our in-depth review.
In part two of our beginners' guides to the fundamentals of cameras, we're going to look at the benefits that a larger sensor can bring. Part three will look at the trade-offs that this brings.
The 2019 Pulitzer Prize photography award winners have been announced in the Breaking News and Feature Photography categories.
French analysts System Plus Consulting have torn down the Huawei P30 Pro and made some interesting findings.
The PhotoCross 15 is the third and largest backpack in ThinkTank's MindShift PhotoCross lineup, rounding out the PhotoCross 10 and 13 variations.
A Kickstarter campaign is aiming to raise funds for the production of an external smartphone flash.
This one-off paper camera makes for great photos, but won't be making any great photos.
In the first of a three-part beginners' guide, we're going to look at whether you need to worry about pixel size when choosing a camera. Parts two and three will look at the benefits of a larger sensor size, and the trade-offs you make.
Nikon's new Z 24-70mm F2.8 S promises a substantial size and weight reduction compared to its F-mount predecessor and a boost in optical quality. See how it performs in our sample gallery.
Self-taught programmer Martin Fitzpatrick has created a Raspberry Pi Zero-powered camera that uses a Pocket Etch-A-Sketch to 'print' the 240x144 pixel image.
Parrot has released an enterprise version of its Anafi drone that features an integrated FLIR thermal camera.
A recent patent application from Canon suggests future pop-up flashes on Canon cameras could have continuous lighting thanks to LEDs.
The Loupedeck+ customizable editing controller now supports Adobe's flagship audio editing program and Apple's professional video editing program.
We spoke to Shigemi Sugimoto, the head of Olympus's imaging division at the CP+ show in Yokohama. He talked to us about the appeal of Micro Four Thirds and gave some hints about the types of technology the company is looking at.
Samyang has announced a new trio of lenses that include an 85mm F1.4 lens for Nikon F-mount systems in addition to a 14mm F2.8 and 85mm F1.4 manual focus lenses for Nikon Z-mount camera systems.
Now that we've finished our full review of Canon's latest affordable full-framer, we've taken a look at how well (or not) it works for some common use-cases.
Andrew Saladino of 'The Royal Ocean Film Society' has shared a humorous, albeit completely fictional look at the history of the camera.
The Ricoh GR series is a perennial staff favorite, and in this week's episode, Chris takes the new GR III to an industrial area, does some street photography in the sunny city, and drinks a tasty beer at a new brewery.
We've been eager to test the Ricoh GR III since it arrived, so in addition to a sample gallery and a thorough DPReview TV analysis, it's also been added to our studio test scene comparison tool. Take a look at how it compares to its peers.
World Press Photo has named John Moore's 'Crying Girl on the Border' the 2019 Photo of the Year and revealed the winners in its other categories. We've rounded up the top singles from each.
Blackmagic has announced a console-inspired keyboard specifically for use alongside its video editing program DaVinci Resolve.
Who said Super 8 cameras were the only cameras that should have multiple lenses you can swap between on the fly?
Samsung has revealed the A80, a triple-camera smartphone with a unique rotating camera mechanism.