Improved Relevance Feedback for Content Based Image Retrieval by Mining User Navigation Patterns

摘要

Content based image retrieval (CBIR) forms the backbone of today's image retrieval systems. Relevance feedback techniques enable accurate results by incorporating user's feedback. However rapid explosion of large scale image databases have resulted in large number of iterative feedbacks from the user to achieve refined search results. This is impractical and inefficient in real applications. A novel method, Navigation-Pattern-based Relevance Feedback (NPRF), is used to achieve the high efficiency and effectiveness of CBIR with the large-scale image data. In terms of efficiency, the iterations of feedback are reduced by using the navigation patterns discovered from the user query log. In terms of effectiveness, the modification in search algorithm NPRFSearch makes use of the discovered navigation patterns and three kinds of query refinement strategies, Query Point Movement (QPM), Query Reweighting (QR), and Query Expansion (QEX), to converge the search space toward the user's intention effectively. By using NPRF method, high quality of image retrieval on RF can be achieved in a small number of feedbacks.