Novel technique for image segmentation based on grammar parsing and Hilbert transform

Abstract

The goal of parsing is to determine whether or not one word belongs to one specified language generated by a grammar. Therefore, the present paper is aimed at exploiting grammar formalism in image processing, especially segmentation. For the English language, the words are those that I am trying to write whereas for an image, the words are the edges and homogeneous regions. We managed to define an image as a set of letters based on image vocabulary, grammar formalism and Hilbert transform for region description. One region is considered as a word recognized by an automaton. We will demonstrate how grammar parsing is practical for synthetic and real images. Indeed, this task is of great help in real time images and video processing with the possibility of taking care of runtime and accuracy challenges. © 2013 Springer-Verlag.