Every day, we recognize hundreds of faces without giving it a second thought. But unlocking exactly how humans pull off that seemingly simple feat has bedeviled researchers, who have been tantalized by the wealth of applications that the science promises. One researcher thinks he has the answer: Our brains are trained to recognize the horizontal “lines” in faces–similar to a barcode.
Dr Steven Dakin of the University College of London manipulated images of the faces of famous people–from George Clooney to Chris Martin, of Coldplay–and found that all the information we need to recognize a face can be reduced to horizontal lines, such as that of the eyebrows and lips, and the sheen of a forehead. He then went further, showing that the information could be reduced to just black and white: A barcode, basically.
Dakin then unlocked another conundrum: What makes us so easily able to tell faces from, say, flowers or trees or any other arbitrary objects. Turns out that in those images, vertical lines tend to predominate. It’s only in faces that all the information becomes reducible to horizontal lines.
Face recognition has been something of a white whale for the tech industry: DARPA, the U.S. Department of Defense’s blue-sky research arm, has long funded research into algorithms that might be used to pick faces out of crowds. Cameras and computer programs such as the new iPhoto are excellent at recognizing when something is a face, and iPhoto in particular allows you to tag a face only once in your photo album to then have that face tagged throughout. But finer grained applications have been elusive, though the promise (and potential misuse) is immense, from homeland security to personalized marketing technology. Dakin may have opened up new paths toward a solution.