Abstract
Visual information retrieval is becoming an important problem in the multimedia age. Currently most image retrieval systems use an image feature vector indexing approach to tackle the problem. In this paper, we present a content-based image retrieval system using the color clusters of the image. Color clustering divides a data set into several groups such that the similarity within a group is larger than that among groups. We present a new image representation using the K-means clustering algorithm and describe a system that applies the color cluster representation and the color cluster as keys to retrieve images. The color cluster representation not only exists inside the image retrieval system, but also presents to end-users as a reference for image retrieval. In the retrieval process, the users can view the color cluster representation of the reference image and select the color clusters of the image in which they are interested. The experimental result shows that the color clustering representation is suitable for feature extraction and intuitive for general user composing query.