Rotation Invariant and Robust Shape Descriptor

for Content-Based Shape Retrieval


S. E. Ouatik1*, A. Elkharraz2, H. Silkan1, M. Oumsis1, A. Lachkar3, M. Meknassi1

1Laboratory of Computer science, Statistics, and quality (LISQ), Faculty of Science Dhar Mahraz., Fès

2National School of Management, Tangier, Morocco

3Laboratory of Electronic, Computer Science and Biotechnology, (LEIB), ESTM, Morocco

* Corresponding author. E-mail: or

Received:12 August 2007; revised version accepted: 22 February 2008



Shape representation and description is one of the major problems in Content-Based Image Retrieval (CBIR). In this paper, we propose an efficient technique based on structural approach for shape representation and description. This latter falls under contour-based techniques, in which the shape boundary is decomposed into parts or tokens. It is an extension of the Berretti et al. shape descriptor, which is not rotation invariant and not robust to noise. In regard to rotation invariance, we propose to normalize the shape orientation by using the concept of major-direction. And after, each token is described with two features: curvature and orientation in a multi-scale representation, this will permit to satisfy the robustness criterion to noise. Finally, in order to test and validate the proposed technique, we have implemented a system for indexing and retrieving of shapes. By using a large shapes database, the performance of our retrieval system was measured in terms of recall and precision. The obtained results highlight the effectiveness of our approach.


Keywords: Structural–based shape boundary representation; Rotation Invariant; Shape Descriptor; Shape Retrieval.

© 2015