Semantic concept and weighted visual feature based image retrieval

The performance of content-based image retrieval is degraded because of the existence of the semantic gap. In order to address this drawback, this paper proposes an image retrieval method fusing with the semantic concept and the weighted visual feature. In the approach, each image is segmented using the normalized cut (N-cut), and the visual characteristics are extracted. After that, the semantic concept is achieved based on the mapping from keywords to image low-level characteristics. The distinctive proportion of the same concept in different images may lead to the priority retrieval of the image which has the same similarity and smaller correlated regions with others. Therefore, we use the weighted visual features to sort the retrieved images. Extensive experiments show the retrieval performance of the proposed method is superior to the traditional content-based image retrieval methods. Copyright © 2014 Binary Information Press.