A color morphology based on pareto-dominance relation and hypervolume measure

Abstract

Challenges in designing non-linear color and multi-spectral image filters can be addressed by known approaches from the field of multi-objective optimization. In particular Pareto-set theory is considered to be feasible. In this paper we explore the impact of so-called indicator functions, which are used to represent all points on a Pareto front. Thereby, we focus on the most common indicator function that is the Hypervolume. It is used to derive an alternative approach to color morphology. We present the conceptional base for the morphological operators, a study of their properties and show their impact in real-world applications, i.e. secure document analysis. © 2008 Society for Imoging Science and Technology.