Histograms analysis for image retrieval

This paper analyzes the use of histograms of low-level image features, such as color and luminance, as descriptors for image-retrieval purposes. A novel definition of histogram capacity curve taking into account the density distribution of histograms in the corresponding spaces is proposed and used to quantify the effectiveness of image descriptors and histogram dissimilarities in image retrieval applications. The results permit the design of scalable image-retrieval systems which make optimal use of computational and storage resources. (C) 2001 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.