Semantic-Based Image Retrieval Using Hierarchical Clustering and Neighbor Graph

In this paper, a data partitioning method was built and applied to a model of semantic-based image retrieval. An improvement of data partitioning based on the hierarchical structure is proposed in which data regions were created to store images; called the growth partition tree (GP-Tree). Based on this, the neighbor cluster graph was built in order to increase the performance of retrieving similar images. The k-NN algorithm was applied to classify an input query image; then, a SPARQL statement structure was generated and executed on the ontology to extract the semantics of an input image as well as the semantics of the image class. From there, the semantic-based image retrieval model was proposed and experimented on the image datasets ImageCLEF, and Stanford Dogs. The experimental results evaluated and compared with the recently published works on the same datasets and demonstrate that the proposed method improves the retrieval accuracy and efficiency.