Relevance Feedback Technology in Content-based Image Retrieval

摘要

Relevance feedback technology provides the method to fill the big gap between low-intermediate features and high-level semantics in image retrieval, thus ensuring the efficiency and effectiveness of contentbased image retrieval.From the three aspects of solving the problems of insufficient feedback, feedback information retrieval device performance and strengthening the real-time, relevance feedback algorithm can be divided into three groups: feedback image selection algorithm, feedback retrieval and sorting algorithm as well as log mining algorithm, etc.This paper puts forward relevance feedback management framework, introduces three kinds of algorithms and thoughts, compares and summarizes them.