Clustering of Invariance Improved Legendre Moment Descriptor for Content Based Image Retrieval

Abstract

This paper reports a k-Means clustering based technique for content based image retrieval (CBIR) using improved Legendre moment descriptor (ILMD). The ILMD is based on orthogonal Legendre moment polynomial and preprocessing steps required for invariance improvement of ILMD is discussed. A comparative study of the clustering accuracy of ILMD with popular Zernike moment descriptor (ZMD) and angular radial transformation descriptor (ARTD) is carried out The clustering accuracy of both contour shape description and region shape description were investigated. The shape databases used for evaluation were MPEG-7 approved CE-1 set B contour shape database and CE-2 set A1 region shape database. The k-Means clustering of the shape descriptors shows better accuracy for ILMD than ARTD and for ARTD than ZMD for both region and contour shape descriptor.