Evaluation of MPEG-7 Shape Descriptors Against Other Shape Descriptors

Dengsheng Zhang and Guojun Lu
Gippsland School of Computing and Information Technology
Monash University
Churchill, Victoria 3842
Australia

Tel: 61-3-9902 6772
Fax: 61-3-9902 6842
Email: dengsheng.zhang, guojun.lu@infotech.monash.edu.au



ABSTRACT

Shape is an important image feature, it is one of the primary low level image features exploited in content-based image retrieval (CBIR). There are generally two types of shape descriptors in the literature: contour-based shape descriptors and region-based shape descriptors. In MPEG-7, curvature scale space descriptor (CSSD) and Zernike moment descriptor (ZMD) have been adopted as contour-based shape descriptor and region-based shape descriptor respectively. In this paper, the two shape descriptors are evaluated against other shape descriptors, the two shape descriptors are also evaluated against each other. Standard methodology is used in the evaluation. Specifically, we use standard databases, large datasets and query sets, commonly used performance measurement and guided principles. A Java based client-server retrieval framework has been implemented to facilitate the evaluation. Results show that Fourier descriptor (FD) outperforms CSSD. CSSD can be replaced by either FD or ZMD.

Keywords: Fourier descriptor, curvature scale space, moments, grid descriptor, CBIR, shape.

full paper