Morphology operators construction by adaptive elliptical structuring elements based on nonlinear structure tensor

Abstract

As the linear structure tensor is tending to inaccurately or even wrong estimate the gradient direction of a gray-level image, we present a novel algorithm to construct adaptive elliptical structuring elements via estimating the local anisotropy of an image based on the nonlinear structure tensor. Erosion, dilation, opening, closing and Hit-or-Miss transform are redefined according to the presented structuring elements, which have been applied to some representative images. The processed results and the quantitative analysis show that the novel morphological operators have more advantages in structure adaptation, corner protection, filtering and targets extraction than the others.