Automatic Extraction of Symmetrical Rectangular Regions from Images using a Genetic Algorithm

Abstract

An image processing method is proposed to automatically extract symmetrical regions from binary and gray-scale images. A symmetrical region is defined as a rectangle whose inner pixel values are almost symmetrical with respect to its center line Position, lengths of two sides and rotational angle of the rectangle are assumed to be unknown In this method, a Genetic Algorithm (GA) is employed to determine the rectangular region In our GA, a set of parameters of a rectangle can be regarded as a chromosome of a virtual living thing. Randomly generated individuals evolve according to natural selection and a good solution corresponding to a symmetrical region is obtained through generation iteration. Arbitrary symmetrical rectangular regions can be extracted by the method without any information about the regions.