Title: Geometric based histograms for shape representation and retrieval
Authors: Nacéra Laiche; Slimane Larabi
Addresses: Computer Science Department, University of Science and Technology, Houari Boumediene, Algiers, Algeria ' Computer Science Department, University of Science and Technology, Houari Boumediene, Algiers, Algeria
Abstract: In this paper, we present a new approach for shape representation and retrieval based on histograms. In the drawback of the proposed histograms descriptor, we consider the concept of curves points. This integration in the proposed histogram-based approach is quite different since geometric description is stored in histograms. The proposed description is not only effective and invariant to geometric transformations and deformations, but also is insensitive to articulations and occluded shapes as it has the advantage of exploring the geometric information of points. The generated histograms are then used to establish matching of shapes by comparing their histograms using dynamic programming. Experimental results of shape retrieval on different kinds of shape databases show the efficiency of the proposed approach when compared with existing shape matching algorithms in literature.
Keywords: log-polar histogram; least squares curve; high curvature points; shape description; shortest augmenting path algorithm; shape retrieval.
DOI: 10.1504/IJAIP.2022.122189
International Journal of Advanced Intelligence Paradigms, 2022 Vol.21 No.3/4, pp.348 - 373
Received: 10 Oct 2016
Accepted: 18 Jun 2017
Published online: 12 Apr 2022 *