An automatic face recognition system has been created that enables machines to identify people from photographs, the science festival was told yesterday.

The system gets over the difficulty people and computers have in identifying unfamiliar faces by emphasising an individual’s essential features.

The software has the potential to help police to pick out terror suspects in a crowd and to make it easier for customs officials to match faces to passport photographs.

The photographs used on identity cards, driving licences and passports could in the future be scrapped and replaced with the composite “essence” pictures.

The system works by creating an “average” picture of a person after dimensions are taken from a series of photographs.

Humans are notorious for finding it difficult to match photographs to strangers’ faces while finding it easy to identify people they know well. The computerised system works in a similar way by creating a composite face from several images of the same person. Once the “average” face has been created the software can pick out that individual’s face from other photographs, even those of poor quality or taken from odd angles.

Experiments with the system, designed by researchers at the University of Glasgow, have shown that it is as successful in correctly identifying familiar faces as humans are.

Perhaps more surprisingly, both humans and the software proved better at identifying a person from the “average” picture than they were from photographs.

Rob Jenkins, who led the development with Professor Mike Burton, said that it took about a dozen photographs for the system to create the average picture of a face.

“The averaging process washes out aspects of the image that are unhelpful, such as lighting effects, while consolidating the aspects of the image that are diagnostic of identity, such as the physical structure of the face,” he said.

“The resulting images are quite uncanny, seeming to bring out the true essence of each face. It also demonstrates that with face recognition, as with so many other problems, we can improve machine performance by mimicking nature’s solution.”

Dr Jenkins said that when the system had been further developed it could play a significant role in catching criminals. He said that the software had the potential to pick out known suspects on closed-circuit television systems.

“This boost in face recognition accuracy has major implications for crime prevention and national security policies,” he told the conference.

However, Dr Jenkins said that how the system might apply to other security uses had yet to be worked out. But he suggested it would be possible within a few years to create average pictures of a person suitable for compulsory use on passports and ID cards.

“This advance is showing a lot of promise at the moment,” he said. “It could be rolled out in a few years if there’s the will to do it.”

Times Recommends

• UK regional accents thrive against the odds
• ‘Scarecrow’ wind farms put rare birds to flight
• What an anti-climax: G-spot is a myth