A New Relevance Feedback Technique For Content-Based Image Retrieval

[[abstract]]相關回饋是一個反覆的過程，讓使用者能根據前次的查詢結果來進行回饋以改善下次的擷取結果。傳統的相關回饋方法屬於短期經驗學習的相關回饋方法，並且無法在多使用者間的多個查詢過程中累積學習經驗。本論文提出了一個新的相關回饋架構用來整合短期及長期的學習經驗，短期的學習經驗由傳統的相關回饋方法而來，而長期的學習經驗由本論文所提的新方法虛擬特徵而來。使用者所有的回饋歷程都會被系統編碼，並表達成有效率的形式，即為虛擬特徵。因此可以根據虛擬特徵來動態的算出影像間的語意相關性。實驗結果是用真實的影像資料庫來進行的；並且結果顯示了本論文所提出的架構比傳統單一查詢過程間作用的相關回饋方法好。[[abstract]]Relevance feedback is an iterative process which refines the retrievals by utilizing the user’s feedback on previously retrieved results. Traditional relevance feedback techniques use solely the short-term learning experience and do not exploit the knowledge collected during cross-sessions with multiple users. In this paper, we propose a relevance feedback framework which combines short-term and long-term learning processes by integrating the traditional methods with a new technique called the virtual feature. The feedback history with all the users is digested by the system and is represented in the virtual features of the images. As such, the dissimilarity measure can be adapted dynamically. The experiments are conducted on a real image database. The results manifest that the proposed framework outperforms the one that adopts a traditional within-session relevance feedback technique.[[note]]碩