IMAGE RETRIEVAL INTERFACE BASED ON VISUALIZATION AND RELEVANCE FEEDBACK

摘要

This paper proposes a new interface mechanism that makes image retrieval user-friendly. The proposed system automatically makes query parameters, visualizes query results, and invites users to establish relevance feedback. This system makes query parameters for image retrieval when the user simply labels some images "Relevant" or "Irrelevant". This system visualizes the retrieval results using Self-Organizing Maps(SOM). This visualization shows system's state which is the interpretation of user's intention, and stimulates the user's feedback process. Some image retrieval tests confirm the effectiveness of the proposed interface.