Semantics oriented approach for image retrieval in low complex scenes

Abstract

The explosive growth of image data leads to the need of research and development of image retrieval. Image retrieval researches are moving from keyword, to visual features and to semantic features. Drive towards semantic features is due to the problem of the keywords which can be very subjective and time consuming as low level features cannot always describe high level concepts in the users’ mind. The main problem encountered in the image retrieval research is the semantic gap that exists between the low-level features and high-level semantics in the images due to the unavailability of low level image features in describing high level concepts in the users’ mind. The aim of this research is to design and validate the semantics oriented approach for image retrieval in low complex scenes. In order to achieve the aim and objectives of the research, the object extraction method for identifying and extracting the objects in a complex scene based on the colour features has been proposed. The semantic extraction and representation method with the semantic image similarity has also been proposed to bridge the semantic gap in image retrieval. In addition, the semantic visual user query, namely Semantic Visual Query Builder (SeVQer), which enables users to express their need and intent at semantic level that reduces the semantic gap in content based image retrieval has been introduced and evaluated. The prototype has been developed to validate the proposed approach in image retrieval. The result of the evaluation shows that the proposed system can achieve the retrieval accuracy of 95.8% and 89.5% for the experiments of semantic object extraction and semantic object and their spatial relationship. The usability evaluation indicated that the proposed semantic visual query achieved higher efficiency and user satisfaction compared to image search by example, keyword and sketch.