Semantic-Based Image Retrieval Using Rs-Tree and Neighbor Graph

In this paper, we present a semantic-based image retrieval method using RS-Tree structure and neighbor graph. The main ideas of the paper include (1) proposing a structure by combining of RS-Tree and neighbor clustering graph, named NBGraphRST; (2) enriching an ontology framework to describe the semantic feature of images. Firstly, a query image is extracted as a low-level visual feature vector and retrieved on NBGraphRST to get a set of content-based similar images. Secondly, the input images are classified by the k-nearest neighbor (k-NN) method to create a set of visual vocabularies. Finally, the SPARQL query is generated to retrieve the semantics of similar images based on ontology. On the basis of the proposed method, the experiment is per-formed on data-sets COREL, Wang, Oxford Flowers-17. The experimental re-sults are compared to related works on the same dataset to demonstrate the ef-fectiveness of the proposed method.