Hermite Transform Based Operators for Orientation Estimation

This paper addresses the problem of the orientation estimation. Basing on the present operator sets dedicated to the orientation estimation, we propose another approach to implement these operators. The approach is based on the Hermite transform and uses the filter functions of certain order to implement the gradient operators and the valleyness operators. The fact shows that neither the gradient operators nor the valleyness ones perform well independently. Both operator sets have their own deficient zones in estimating orientation. However, the deficient zones of each operator set are different and exclusive. After conducting the analysis towards the orientations estimated by each operator set, we find a way that makes them cooperate. Finally we use the synthesized images and the real ones to test the competence of the operator sets. The results indicate their good performance on orientation estimation.