Appling grey relational analysis to the relevance feedback in content-based image retrieval

Based on the quantitative grey relational analysis method, a simple and effective user query learning algorithm for the relevance feedback in content-based image retrieval is proposed. This new approach is an supervised algorithm, and the query parameters can be dynamically updated via relevance feedback to reflect the user´s particular information need. Experimental results shows that the proposed method performs better than the previous GRA-based algorithms for learning the query parameters in the learning precision and the generalization ability, and thus the performance of the relevance feedback for content-based image retrieval can be considerably improved.