July 9, 2012
Taking Nothing At Face Value
Photographs of faces may not be adequate proof of a person's identity and this could have serious implications for the accuracy of passport photographs in determining identity. Research funded by the Economic and Social Research Council (ESRC) shows that an image of a person may look strikingly different from one image to the next. We are told not to smile in our passport photos as a smile distorts the face; but the opposite may actually be true, and a poker face may be the one which distorts normal facial features.
Dr Rob Jenkins and his team at the University of Glasgow took a sample of photos from the internet to show the wide range of differing images of one person. In a series of experiments, viewers unfamiliar with the subject of the photograph believed that the photos they were viewing were of different people — when in fact they were simply different presentations of the same person.
Further experiments looked at the perceived attractiveness of an image. They showed that variability within a person was greater than the variability between people when it came to deeming a face attractive. The experiments showed that faces and facial photographs cannot be considered to be representative of each other. Facial recognition must start to consider not only how to tell people apart, but also how to spot the similarities.
Dr Jenkins states: "The sheer variation in photos of an individual´s face did bring us up short. Previous research on identification has focused on differences between faces. Now it turns out that differences within faces are just as large. Therefore in this study we have discovered a new dimension to the field of face recognition."
Assessing a face as attractive can have positive consequences when it comes to finding a mate, finding a job and perhaps seeking approval. So it is important to choose the best image of ourselves to present to a judgmental world. Dr Jenkins states: "This research makes us consider much more deeply what it means to have a 'good likeness'".
On the Net: