Clustering of Texture Features for Content-Based Image Retrieval

Content-based image retrieval has received significant attention in recent years and many image retrieval systems have been developed based on image contents. In such systems, the well-known features to describe an image content are color, shape and texture. In this paper, we have studied an approach based on clustering of the texture features, aiming both to improve the retrieval performance and to allow users to express their queries easily. To do this, the texture features extracted from images are grouped according to their similarities and then one of them is chosen as a representative for each group. These representatives are then given to users to express their query. Besides the detailed descriptions of clustering process and a summary of results obtained from the experiments, a comparison about statistical texture extraction methods and effects of clustering to them are also presented.