Detection of symmetric features in images

Features detection plays a significant role in many computer vision tasks. There exist a plenty of various methods for features detection in an image processing domain. Symmetric features form a substantial class of them which takes an advantage of robust approach and fast implementation. In this paper a basic concept of a mirror and rotational symmetry is introduced and several implementation aspects are discussed. Symmetry detection process is transparently compared to the selected standard approach and results are demonstrated on the rotational symmetry task.

I. Introduction

DETECTION of features in images is one of the most important and the most frequently used activity in order to understand content of an image. There was performed a significant research particularly on the field of so-called point of interest in the past [1], [2]. Several other concepts have emerged from time to time when better performance or higher accuracy was required [3]. The recent research attention has been paid for so-called symmetry constellations, which gives a new idea about detection of symmetrical points of interest [4], [5]. The basic concept of symmetry features detection is introduced in this paper besides implementation issues which are described at the end of the paper.