Curvature estimation in oriented patterns using curvilinear models appliedto gradient vector fields

Curved oriented patterns are dominated by high frequencies and exhibit zero gradients on ridges and valleys. Existing curvature estimators fail here. The characterization of curved oriented patterns based on translation invar iance lacks an estimation of local curvature and yields a biased curvature- dependent confidence measure. Using parameterized curvilinear models we mea sure the amount of local gradient energy along the model gradient as a func tion of model curvature. Minimizing the residual energy yields a closed-for m solution for the local curvature estimate and the corresponding confidenc e measure. We show that simple curvilinear models are applicable in the ana lysis of a wide variety of curved oriented patterns.