On Points Geometry for Fast Digital Image Segmentation

Abstract

Automatic segmentation of digital images is of utmost importance in many applications especially those related to machine vision, i.e., robotics and vision in the production industry. There exist different algorithms to carry out image segmentation. This paper presents a novel image segmentation algorithm with low computational complexity. The proposed approach is based on Delaunay triangulation (DT) which is the dual of Voronoi Diagram (VD), a well-known technique in computational geometry. VD is presently implemented in many areas, but researchers primarily focus on its use in skeletonization and in generating Euclidean distances. DT Triangles form a mesh from any input dot patterns which are in our case the image intensity colours. The union of all triangles is called the convex hull which helps generate clusters of intensity values using information from the vertices of its external boundary. In this way, it is possible to produce segmented image regions. We used the BioID face database to test our algorithm as we aim to produce an adaptive Steganography system which uses an object oriented approach.