U. SOUTHAMPTON (UK) — A technique that highlights tubular structures like ears is making it possible for scientists to automatically identify people.
The method uses light rays to highlight circular and tubular structures, such as the curved rim at the top of the ear, known as the helix, or glasses frames.
By extracting the elliptical shape of the helix, it can be used as the basis of a method for the discovery, localization, and normalization of the image for ear biometrics.
Mark Nixon, professor of electronics and computer science at the University of Southampton first proved that ears were a viable biometric in 2005.
At the time he said that ears have certain advantages over the more established biometrics, such as face recognition, as they have a rich and stable structure that is preserved from birth to old age, and instead of ageing they just get bigger.
The ear is not affected by changes in facial expression and remains fixed in the middle of the side of the head against a predictable background, unlike face recognition which usually requires the face to be captured against a controlled background.
Because ears can be concealed by hair, Nixon came up with new algorithms to make it possible to identify and isolate the ear from the head.
The new technique achieved 99.6 percent success at identifying ears from more than 250 images, despite hair concealment and possible confusion with glasses. The results show great potential for enhancing the detection of structural features.
“Feature recognition is one of the biggest challenges of computer vision,” Nixon says. “The (new) technique may also be appropriate for use in gait biometrics, as legs act as tubular features that the transform is adept at extracting.
“The transform could also be extended to work upon 3-D images, both spatial and spatio-temporal, for 3-D biometrics or object tracking. As a general pre-processing technique for feature extraction in computer images, the technology is now pervading manufacturing, surveillance, and medical applications.”
Nixon’s paper was presented at the IEEE Fourth International Conference on Biometrics: Theory, Applications and Systems, held recently in Washington, D.C.
More news from the University of Southampton: http://www.soton.ac.uk/