An Approach to Content-Based Image Retrieval using Clustering and Space Transformation

In this paper, we propose a new approach to content-based image retrieval that uses space transformation methods to transform the original low-level image space into a reduced dimension highlevel vector space that enables efficient query processing. Our retrieval system consists of two phases: database population and image retrieval. Our approach is generic as it can be applied to many other application domains, in addition to image retrieval. Our approach is efficient, since we use the inexpensive “estimated” distance, as opposed to the computationally inefficient “real” distance, to retrieve the relevant results for a given query image. Our results show that the space transformation methods that we compared preserve either the distances or the order of the distances in the transformed space.