Multiscale local multiple orientation estimation using mathematical morphology and B-spline interpolation

This paper introduces a novel multiscale approach to estimate local multiple orientations, which are underlying in the discrete grid of an image. The basic ingredients are two: on the one hand, multiscale directional openings by line segments of variable length, which produce directional signatures for various scales, with the estimation of the orientation properties; and on the other hand, multiple peak detection by means of b-spline interpolation of the directional signatures. Experimental results on real and synthetic images show the applications of the proposed method based on mathematical morphology, which can detect local multiple orientations in textured images at different scales with high accuracy.