Varying neighbourhood parameters for the computation of normals on surfaces in discrete space

Abstract

A method is examined which computes normal vectors on binary discrete surfaces by making use of the local separation criterion of surfaces. The method considers points in the neighbourhood of a surface point in object space to recover information about the local surface inclination. The article examines and evaluates the different parameters used in the computation of the normals, such as neighbourhood size and weighting factors, both from a numerical and from a visual point of view. Moreover far the discrete space the use of spherical neighbourhoods is introduced for the computation, and the vectors computed from the two different sides of a surface are compared.