Smile: you’re on camera!

Font Size

Print

Feedback

Email

Discuss

: Does that new feature increasingly found in pocket-sized digital cameras—face-recognition technology—really work? It’s actually a lot cleverer than you think. A few years ago, it would have needed a shoe-box of electronics to drive it, and it would still have been hit-or-miss. But in the brutally competitive world of digital photography, Canon, Pentax and Fuji have honed the technology so their popular digicams can take more striking pictures by finding, and then focusing on, the faces in the viewfinder.

With 70% of all photographs being taken primarily of people, much is to be gained from using a face-recognition algorithm stored in the camera’s chip. This scans the image in the viewfinder for a shape resembling a human face—ie, eyes, ears, nose and chin. Once located, the camera can then adjust the focus exclusively for that part of the picture. Some cameras can recognise up to ten faces in a scene and set an average focus, or select just one face from a group and focus on that.

To prevent the camera from locking on to faces in the background, the algorithms used in today’s digicams tend to ignore features smaller than 10% of the viewfinder’s height. The result is a pin-sharp image of the subject’s facial features—the part we’re interested in—amid a slightly blurrier foreground and background. Niftier still, the algorithm can also capture the face’s actual location within the scene. That lets the user zoom in automatically on the face immediately after the picture has been taken, to check everything is okay before saving it.

Now face-recognition technology is getting even smarter. Sony is due to launch a digital camera that can be set so it won’t release the shutter until people in the picture are smiling. The software analyses the scene for facial expressions associated with happiness—including the upturn of the corners of the mouth, the separation of the lips, and the wrinkle of the eyes. You can designate which of up to eight people in the viewfinder to focus on, and select three different facial expressions: smile, grin or laugh. Even so, this is still a long way from proper face recognition. Despite what the camera makers say, such techniques merely detect faces; they do not actually recognise them.