An enhanced shape descriptor based on radial distances

Abstract

Contour and region based descriptors are among the main approaches in content based image retrieval (CBIR). In the contour based approaches, Fourier Transform is applied on the shape signatures to obtain Fourier descriptors (FD), which has attractive properties such as rotational, scale, and translational invariance. However, the FD derived from different signatures can have significant effect on the performance of retrieval. In this paper we introduce a new shape signature, which is called the radial distance with triangular centroid area (RDTCA) signature. Based on our retrieval experiments using Kimia 99 and Kimia 216 dataset, FD derived from RDTCA signature outperformed the FD derived from other signatures in terms of robustness and retrieval performance. © 2011 IEEE.