Image Retrieval with Long-Term Memory Learning and Short-Time Relevance Feedback

This paper presents a framework for effective image retrieval by employing long-term memory learning together with short time relevance feedback. Considering the fuzzy characteristic of the human's semantic knowledge, the proposed retrieval framework is associated with a novel matrix named fuzzy semantic relevance matrix (FSRM), which is used for remembering the semantic information obtained from user-provided interactions. Experimental results demonstrate the effectiveness of the proposed scheme.