A New Algorithm for Greyscale Objects Representation by Means of the Polar Transform and Vertical and Horizontal Projections

Abstract

Intelligent computer vision systems can be based on various approaches, methods and algorithms. One of them is the usage of object descriptors, devoted to the representation of objects extracted from digital images (or video sequences) mainly by means of low-level features. There are several features that are applied in computer vision theory and practice, e.g. color, context of the information, luminance, movement, shape, and texture. Amongst them shape, color and texture are especially popular. To the contrary, object representation based on greyscale is less popular. The paper proposes and analyses a new simple and fast algorithm for greyscale object representation. The description method is based on the usage of the polar transform of pixels belonging to an object, and projections – vertical and horizontal. Apart from these operations, some additional steps are also applied in order to improve the efficiency of the developed approach, e.g. median and low-pass filtering. The properties of the proposed greyscale object representation algorithm are analyzed experimentally by means of an exemplary application from the computer vision domain. The ear images (taken from The West Pomeranian University of Technology Ear Database) are applied in the experiment. The obtained results constitute the basis for certain conclusions as well as the proposition of future plans and works on the problem.