Mathematical Morphology in the Linguistic HSI Color Space

PASTORE J. I.; ESPIN ANDRADE R.; BALLARIN, V.

Ciudad de Juarez

Workshop; International Virtual Workshop of Business Analytics; 2019

EUREKA and Autonomous University of Ciudad Juarez (UACJ)

Color is a very important visual feature used in computer vision and image processing. Compared with grayscale images, color images can provide richer information. However, the direct extension of gray scale image algorithms to color is not always straightforward. Mathematical Morphology (MM) is founded on lattice theory, therefore the most elementary requirement to define morphological operators is to establish an ordering of the space of the pixel intensities. That is why the current extension to color approaches of the classical MM focus on the election of a lattice model for the color space.The use of fuzzy set theory is appropriate to manage the imprecision in color description. Moreover, in practical applications it is usual to work with different color terms, whose number and design depend on the application itself. In this sense, the concept of linguistic color space is useful, among other things, for representing the set of fuzzy colors that are relevant to a certain application.Several attempts have been made and different approaches have been presented in the last years, aiming at building a fuzzy mathematical morphology model. The situation become more complex when trying to apply fuzzy set theory in color images because of the existence of many different ordering schemes and different definitions for the basic morphological operators. In this paper, we propose a novel approach using linguistic variables to express fuzzy predicates in HSI color space in order to define the fundamental morphological operations of erosion and dilation. This proposal allows to reduce the ambiguity in color description and avoid false colors.