Skewness balancing algorithm for approximation of discrete objects boundaries

Object boundary is an important feature for image processing and computer vision applications. In this paper a new method for extracting the non convex boundaries of an object represented by 2D point clouds is established. In order to determine the object boundaries we started by constructing the convex-hull-based Delaunay triangulation using the point clouds. Given the fact that the points are sampled from the object surface using an instrument such as cameras or laser scanners, the distribution of the edges lengths belonging to the objects follows a Gaussian distribution. However this distribution is skewed due to the existence of long edges introduced by the Delaunay triangulation. Removing the skewness will make the convex boundary built by the Delauny algorithm converge to the real boundary of the object. We tested our method using different datasets that includes synthetic data, urban LiDAR (Light Detection and Ranging) data, and binary images. The results show that the proposed method successfully extracts the object boundary.