Abstract:

Relevance feedback has been proposed as an important technique to boost the retrieval performance in contentbased image retrieval (CBIR).However,since there exists a semantic gap between lowlevel features and high level semantic concepts in CBIR, typical relevance feedback techniques need to perform mutiple iterations of feedback for achieving satisfactory results. These procedures are timeconsuming and may make the users bored in the retrieval tasks. Therefore, this paper proposes a novel feedback strategy based on analysising the usage and storage process of the retrieval feedback historical information in contentbased image retrieval. The testing results of the prototype system show that the strategy we put forward significantly improves the efficiency of retrieval performance and performs better than typical relevance feedback strategies.

Key words: relevance feedback;collaborative filtering;buddy image;contentbased image rentrieval