Dissimilarity measures for content-based image retrieval.
MetadataShow full item record
Dissimilarity measurement plays a crucial role in contentbased image retrieval. In this paper, sixteen core dissimilarity measures are introduced and evaluated. We carry out a systematic performance comparison on three image collections, including Corel, Getty and Trecvid2003, with seven different feature spaces. Two search scenarios are considered: single image queries based on Vector-Space-Model, and multiimage queries based on k-Nearest Neighbours search. A number of observations is drawn, which will lay a foundation for developing more effective image search technologies.