A semantic-based image retrieval system using nested KD-Tree structure

Abstract. Semantic-based image retrieval system has relied on many different methods. In this paper, a semantic-based image retrieval method using Nested KD-Tree structure and ontology is proposed. The Nested KD-Tree structure is built depending on an image classification method based on KD-Tree. Firstly, KD-Tree is built as a balanced multi-branch tree; Secondly, some leaf nodes on KD-Tree are grown into SubTrees to create a Nested KD-Tree structure. Therefore, an algorithm of Nested KD-Tree construction and a method of image classification are performed. Each query image is extracted to a feature vector to perform image classification and retrieve similar images based on Nested KD-Tree. Then, a SPARQL query is generated to retrieve similar images by semantic-based on ontology. On the basis of this theory, a semantic-based image retrieval model is proposed to build an experiment. The precision on experimental image data sets including COREL, Wang, and Caltech101 of 81.19%, 80.29%, 72.55%, respectively. The experiment results are compared to the published works on the same data set to demonstrate the efficiency of our proposed method.