A search engine that uses sophisticated facial recognition to allow users to identify and find people in online images will launch next month. But civil liberties groups say the biometric-style tool could compromise the privacy of anyone who has their picture online.
Search engine Polar Rose reconstructs the 3D shape of a person's face and then combines that with characteristics of their features to generate a unique "face print". This can then be used to search other photos for a match.
In January users will be able to download a plugin for their browser that allows users to enter information about faces they recognise in online images. This data is then sent to a central server allowing anyone looking at an image containing that particular face print to tell who it is. Users can also search the web for more photos containing that face.
Online image search engines usually work much like their text counterparts. "They find images on pages that contain the words you search for," says Jan Erik Solem, whose PhD project at Malmö University College, Sweden, led to the new company. "Search engines are blind to images, Polar Rose is not."
"Some biometric companies are using 3D laser scans of faces to aid identification from photos," Solem says. "We've developed a way to work backwards; we can create a 3D model of a face from a 2D image."
That allows Polar Rose to recognise people even when the pose or lighting has changed, he says. The technique was developed using a database of around two thousand 3D face-scans paired with normal 2D photos.
"We used statistical methods to work out the relationship between the two," explains Solem.