EFFICIENT BLOCK-BASED MATCHING FOR CONTENT-BASED IMAGE RETRIEVAL USING COLOR FEATURES

摘要

This paper presents a simple but reliable block-based approach to address region-based image retrieval using color features. The proposed algorithm integrates a block-based localization technique, which offers flexible feature vector property for block matching purpose, together with several histogram-based color features, namely the color histogram, color coherence vector and color co-occurrence histogram. The algorithm was tested on 2250 images of small objects from the Amsterdam Library of Object Images (ALOI) database, quantitatively and qualitatively and promising results are reported.