Image indexing using moments and wavelets

Abstract

Histogram comparison is a popular technique for image and video indexing. The complexity of the technique can be reduced by representing the histogram by its moments. In this paper, we propose two techniques to improve the performance of the basic histogram/moment-based technique. First, we propose to use orthogonal Legendre moments for representing histograms. Since Legendre moments are orthogonal, they provide superior indexing performance compared to regular moments at a similar complexity. Secondly, we propose to compare the histograms of wavelet coefficients at different scales. The wavelet coefficients provide important directional information, and hence improve the performance of the basic histogram-based technique. The proposed scheme can be easily extended to color images and also be integrated into a wavelet-based image coder.