Long-Term Learning in Content-Based Image Retrieval

摘要

In content-based image retrieval, relevance feedback is an interactive process, which builds a bridge to connect users with the search engine. It leads to much improved retrieval performance by updating the query and the similarity measure according to a user's preference; and recently techniques have matured to some extent. However, most previous relevance feedback approaches exploit short-term learning (intraquery learning) that is dealing with the current feedback session but ignoring historical data from other users, which potentially results in a great loss of useful information. Fortunately, by recording and collecting feedback knowledge from different users over a variety of query sessions, long-term learning (interquery learning) can be implemented to further improve the performance of content-based image retrieval in terms of effectiveness and efficiency. For this reason, long-term learning has an increasingly important role in multimedia information searching. No comprehensive survey of long-term learning has been conducted to date. To this end, the article addresses this omission and offers suggestions for future work.