Novel Relevance Feedback Approach to Content-Based Image Retrieval

Abstract

In content-based image retrieval systems, relevance feedback is an important way to narrow the gap between low-level features and high-level conception. Relevant images were often considered in traditional relevant feedback. But irrelevant images weren't considered. The method of moving query vector and method of updating weight factor are only separately used in content-based image retrieval. A new approach is proposed by combining moving query vector with updating weight and the influence of both relevant and irrelevant images are considered for image retrieval. The results demonstrate that retrieval performance is obviously improved. It is positive meaning for the development of relevance feedback.