Visual information retrieval in images and video has been developing rapidly in our daily life and is an important research field in content-based information indexing and retrieval, automatic annotation and structuring of images. Visual information system can make the use of relevance feedback so that the user progressively refines the search result by marking images in the result as relevant , not relevant or neutral to the search query and then repeating the search with the new information. With a comprehensive review as the main portion, this paper also suggested some novel solutions and perspectives throughout the discussion. Introduce the concept of Negative bootstrap, opens up interesting avenues for future research.