A novel set of Bilateral and Radial Symmetry Shape Descriptor based on contour information

NATALIA V. REVOLLO; CLAUDIO A. DELRIEUX; ROLANDO GONZALES-JOSE

IET COMPUTER VISION

INST ENGINEERING TECHNOLOGY-IET

Lugar: Londres; Año: 2016

1751-9632

Form and shape descriptors are among the most useful features for object identification and 12 recognition. Even though there exists a widely used set of shape descriptors and underlying 13 computational methods for their evaluation, such descriptors are usually unable to distinguish 14 among very similar objects, notwithstanding that they may appear very different to the human 15 eye. We propose a new set of shape descriptors based on a finer determination of the object main 16 axes and a more accurate estimation of their bilateral and radial symmetry. These descriptors 17 are thoroughly tested using several synthetic objects with varying degrees of symmetry. The 18 methods for axes estimation and symmetry descriptors extraction outperform the widespread 19 shape descriptors in recognising and identifying among very similar objects. Also, our proposal 20 is more robust in recognition under scaling and rotation.