“Ears are a particularly appealing approach to noncontact biometrics because they are relatively constant over a person’s life and are unaffected by expressions, unlike faces.”
– explain researchers from the Image Processing and Computer Vision dept. at the University of Southampton, UK in a recent article for IEEE Transactions on Systems, Man and Cybernetics (A). In which professor Mark S Nixon BSc PhD FIAPR FIET CEng and colleague John D. Bustard (Ph.D researcher) draw attention to the large potential of ears for biometric recognition purposes. But although electronic ear-recognition systems have been under investigation for more than a decade, reliable computerised recognition of individual ears is not entirely straightforward.
“In particular, ears have to be recognized from different
angles (poses), under different lighting conditions, and with different cameras. It must also be possible to distinguish ears from background clutter and identify them when partly occluded by hair, hats, or other objects.”
The Southampton team have developed a new technique using scale-invariant feature transform and homographies calculated from SIFT point matches. It can cope not only with fuzzy and degraded images, but also with ears that are up to 18% occluded. There is further work to be done however, before fast and reliable ear-recogition becomes a reality – in particular
“To fully automate the enrolment process, there is a need to construct a model of ear variation so that novel ears can be detected”
For full details, the paper can be examined here :
Note: Sample ears for the study were found at the Extended Multi Modal Verification for Teleservices and Security applications database (XM2VTSDB) at the University of Surrey.