Thinking of a System for Image Retrieval

Increasing applications are demanding effective and efficient support to perform retrieval in large collections of digital images. The work presented here is an early stage research focusing on the integration between text-based and contentbased image retrieval. The main objective is to find a valid solution to the problem of reducing the so called semantic gap, i.e. the lack of coincidence existing between the visual information contained in an image and the interpretation that a user can give of it. To address the semantic gap problem, we intend to use a combination of several approaches. Firstly, a linking between low-level features and text description is obtained by a semi-automatic annotation process, which makes use of shape prototypes generated by clustering. Precisely, the system indexes objects based on shape and groups them into a set of clusters, with each cluster represented by a prototype. Then, a taxonomy of objects that are described by both visual ontologies and textual features is attached to prototypes, by forming a visual description of a subset of the objects. The paper outlines the architecture of the system and describes briefly algorithms underpinning the proposed approach.