We propose a new approach to skew-symmetry detection, based on the the
ory of invariant signatures for planar objects. Invariant signatures a
ssociated to object boundaries are generalizations of the curvature ve
rsus arclength description of curves, invariant under geometric transf
ormations more complex than the Euclidean ones. We show that symmetrie
s of objects, and hence of closed boundaries, translate into simple st
ructures in the invariant signature functions and are therefore, in pr
inciple, readily detectable. (C) 1997 Pattern Recognition Society. Pub
lished by Elsevier Science Ltd.