A color-based clustering approach for web image search results
Guangqing Liu, Byung-Wook Lee · 2009
In this paper, we propose an approach to cluster web image search results based on color (CbC). We index images into color histogram descriptors and improved-color distribution entropy descriptors (I-CDE), and cluster them by the similarity of color descriptors. However, there are some problems in the similarity measurement of I-CDE. In order to resolve the problems, we introduce a new method that measures similarity in vector space model (VSM). Through experiment and comparison analysis, it is proved that the proposed CbC approach supplies a good experience for users, and the VSM method solves the problems existing in the similarity measuring method of I-CDE and gives better performance.