An Object Properties Filter for Multi-Modality Ontology Semantic Image Retrieval

Abstract

Ontology is a semantic technology that provides the possible approach to bridge the issue on semantic gap in image retrieval between low-level visual features and high-level human semantic. The semantic gap occurs when there is a discrepancy between the information that is extracted from visual data and the text description. In other words, there is a difference between the computational representation in machine and human natural language. In this paper, an ontology has been utilized to reduce the semantic gap by developing a multi-modality ontology image retrieval with the enhancement of a retrieval mechanism by using the object properties filter. To achieve this, a multi-modality ontology semantic image framework was proposed, comprising of four main components which were resource identification, information extraction, knowledge-based construction and retrieval mechanism. A new approach, namely object properties filter is proposed by customizing the semantic image retrieval algorithm and the graphical user interface to facilitate the user to engage with the machine i.e. computers, in order to enhance the retrieval performance. The experiment results showed that the proposed approach delivered better results compared to the approach that did not use the object properties filter based on probability precision measurement.