July 31, 2011
Artificial photos: A protection algorithm
A trio of computer scientists have conjured up a photography breakthrough. It involves a new piece of software that tricks the viewer into believing a photo was shot from a different angle. It might not sound like a monumental change, but it’s a photography breakthrough that could one day save lives.
Back in 2007, computer scientists Peter Schaffer, Djamila Aouada and Shishir Nagaraja were shocked when photographers at the 2007 Burmese protests were identified and arrested by government officials. They had been fingered by their own photography.
The officials were able to deduce their identities by pinpointing the location they took the photos using CCTV. A cross-check between the photographer’s location and a range of snaps revealed all.
What the photographers needed was a way to report from the field, without fear of being identified and targeted by hostile forces afterwards. Schaffer, Aouada and Nagaraja set to work.
Their answer is a smart piece of software that grabs photos from a camera in order to make them appear is if they were shot from a different vantage point.
How does it work?
Photographers must take several photos of a scene for the technique to work properly, although it’s able to function with just two original images.
Those photos are downloaded from the camera’s memory card, before being run through “view synthesis algorithms” that determine changes that can be made to an image, subtly adjusting angles to obfuscate the photographer’s position.
The software analyses the images, or “frames”, looking for corresponding objects and then merges them together. The result is an image with a new viewpoint, which is an amalgamation of the originals.
“The photographer’s anonymity can hopefully be maximized to the entire set of people present in that space,” explain the software’s inventors.
The software has yet to be tested in the field, but already there are a multitude of ways it could be used. The possibilities of reporting without fear in such situations are only likely to increase the quality and quantity of coverage, while undercover reporters tasked with infiltrating criminal organisations or companies will no doubt welcome it too.
The camera never lies, we’re told, but with technology on its side maybe soon it’ll start to bend the truth a little.
Read Peter Schaffer, Djamila Aouada and Shishir Nagaraja’s research paper in full here: [PDF link].