Download:

Biometrics with direction

Prof. Josef Bigun
Halmstad University and Chalmers University of Technology, Sweden

Face is an important trait with which humans recognize and identify other humans. There is significant evidence, both from studies of prosopagnosia and from studies of brain damage, that face analysis engages special signal processing in visual cortex that is different from processing of other objects [1, 4, 5]. There is a general agreement that approximately at the age of 12, the performance of children in face recognition reaches adult levels, that there is already an impressive face recognition ability by the age of 5, and that measurable preferences for face stimuli exist in babies even younger than 10 minutes [3]. For example, human infants a few minutes of age show a preference to track a human face farther than other moving nonface objects [7]. While there is a reasonable rotation-invariance to recognize objects though it takes longer times, turning a face upside down results usually in a dramatic reduction of face identification [2]. These and other findings indicate that face recognition develops earlier than other object recognition skills, and that it is much more direction sensitive than recognition of other objects.

Perhaps recognition of face identities is so complex that encoding the diversity of faces demands much more from our general-purpose, local direction, and frequency-based feature extraction system. If so, that would explain our extreme directional sensitivity in face recognition. Parallely, there is mounting evidence that faces [1, 6, 8, 9], just like color, disposes its own "brain center". Face sensitive cells have been found in several parts of the visual cortex of monkeys, although they are found in most significant numbers in a subdivision of inferotemporal cortex in the vicinity of the superior temporal sulcus. Whether these cells are actually necessary and sucient to establish the identity of a face, or if they are only needed for gaze-invariant general human face recognition (without person identity) is not known to sucient accuracy. In humans, by using magneto resonance studies, the face identity establishing system engages a brain region called fusiform gyrus. However, it may not exclusively be devoted to face identification, as other sub-categorization of object tasks activate this region too. On the other side, directional processing of images is the most ubiquous signal processing of the known mammalian vision. It feeds nearly all other visual processing areas and subsystems, including face recognition/spotting [hubel,orban,zeki,baylis]. This means that when humans act as supervisors or operators of even other biometrics data that are images, such as fingerprints and iris, they rely on directional processing.

It is therefore not surprising that many of machine biometrics methods that rely on directional features are among the most successfull, regardless the trait, e.g. face recognition, fingerprint recognition, iris recognition. In 2-D, the earliest solutions to the problem of finding direction of an image patch, consist in projecting the image onto a number of fixed orthogonal functions. The thus obtained projection coeficients are used to derive the orientation parameter of the model. When the used number of filters is increased, the local image is described increasingly better but the inverse mapping of the projections to an optimal direction increases in complexity rapidly. Here we will present an approach that models the shapes of iso-curves of images via direction tensor fields. The concept offers a unified theory to both Gabor filtering based direction tensor field estimation and

Gaussian derivatives based direction tensor field estimations, both being among the most popular features used in biometric authentication, e.g. face, fingerprint, and iris. Examples illustrating the theory will be detailed.

References

[1] G.C. Baylis, E.T. Rolls, and C.M. Leonard. Functional divisions of the temporal lobe neocortex. J. Neuroscience, 7:330–342, 1987.

[2] V. Bruce and A. Young. In the eye of the beholder. Oxford University Press, Oxford, 1998.

[3] H.D. Ellis, D.M. Ellis, and J.A. Hosie. Priming effects in children’s face recognition. British Journal of Psychology, 84:101–110, 1993.

[4] M.J. Farah. Is face recognition special? evidence from neuropsychology. Behavioral Brain Research, 76:181–189, 1996.

[5] I. Gauthier and M.J. Tarr. Becoming a "greeble" expert: Exploring mechanisms for face recognition. Vision Research, 37:1673–1682, 1997.

[6] M.E. Hasselmo, E.T. Rolls, G.C. Baylis, and V. Nalwa. Object-centered encoding by face-selective neurons in the cortex in the superior temporal sulcus of the monkey. Experimental Brain Research, 75:417–429, 1989.

[7] M.H. Johnson. Developmental cognitive neuroscience. Blackwell, 1997.

[8] D.I. Perrett, P.A. Smith, D.D. Potter, A. Mistlin, A.S. Head, A.D. Milner, and M.A. Jeeves. Visual cells in the temporal cortex sensitive to face view and gaze direction. Proc. of the Royal Society of London, Series B, 223:293–317, 1985.

[9] S. Yamane, S. Kaji, and K. Kawano. What facial features activate face neurons in the inferotemporal cortex of the monkey. Experimental Brain Research, 73:209–214, 1988.

with support from