H. Zha, R.-i. Taniguchi, and S. Maybank (Eds.): ACCV 2009, Part II, LNCS 5995, pp. 448–457, 2010. © Springer-Verlag Berlin Heidelberg 2010 Region Based Color Image Retrieval Using Curvelet Transform Md. Monirul Islam, Dengsheng Zhang, and Guojun Lu Gippsland School of Information Technology, Monash University, VIC 3842, Australia {md.monirul.islam,dengsheng.zhang, guojun.lu}@infotech.monash.edu.au Abstract. Region based image retrieval has received significant attention from recent researches because it can provide local description of images, object based query, and semantic learning. In this paper, we apply curvelet transform to region based retrieval of color images. The curvelet transform has shown promising result in image de-noising, character recognition, and texture image retrieval. However, curvelet feature extraction for segmented regions is chal- lenging because it requires regular (e.g., rectangular) shape images or regions, while segmented regions are usually irregular. An efficient method is proposed to convert irregular regions to regular regions. Discrete curvelet transform can then be applied on these regular shape regions. Experimental results and analy- ses show the effectiveness of the proposed shape transform method. We also show the curvelet feature extracted from the transformed regions outperforms the widely used Gabor features in retrieving natural color images. 1 Introduction Regions are fundamental blocks for recent region based image retrieval (RBIR) tech- niques involving regions and semantic learning [1]. Texture feature is an essential component in most region based image retrieval (RBIR) techniques because of its strong discriminative power. Many texture feature extraction techniques have been proposed including spatial and spectral. Spatial techniques are subject to noise and difficult to obtain. So far, spectral methods, like Gabor [2, 3] and wavelet [4], have shown the best retrieval performance. Recent researches show that curvelet transform has significant advantages over Gabor due to curvelet is more effective in capturing curvilinear properties, like lines and edges [5]. It shows promising results in image de-noising [6], character recognition [7] and texture image retrieval [8]. To date, no application has been reported on real world image retrieval using curvelet transform. This paper applies curvelet transform in a region based image retrieval technique to retrieve color images. The application of curvelet transform in a region based tech- nique is challenging due to the fact that curvelet transform requires rectangular im- ages or regions, while segmented regions are usually irregular, as shown in Fig. 1. Most of the existing RBIR techniques define a region as a set of small blocks of size 4 by 4 pixels and apply spectral transform on those blocks [9]. Then the feature of the region is calculated as the average feature of those blocks. This technique has