Linguistic Expression Based Image Description Framework and its Application to Image Retrieval

Abstract

Content-based image retrieval (CBIR) has been researched for many years, but there are quite a few problems affecting the development of the CBIR. Most of the disappointments with early CBIR systems come from the lack of recognizing the existence of the semantic gap and its consequences for system set-up. The semantic gap is the lack of coincidence between the information that one can extract from the visual data and the semantics that the same data have for a user in a given situation. Systematically describing and automatically obtaining the image semantic features are open issues [1]. We are not based on the digital feature vectors, but on the concepts describing the target images in natural language when we search images in the database or internet. But natural language is imbued with imprecision, vagueness and other forms of uncertainty in its syntactic structure and semantic content. It becomes a difficult task for computers to describe images with natural or natural-like language. How to utilize languages to describe the semantic features of an image becomes a significant but difficult problem. In this chapter, we propose a linguistic expression-based image description (LEBID) framework to provide images with semantic expression based on linguistic variable in fuzzy theory [2]. LEBID systematically demonstrates how to depict the image semantic features with natural language based on linguistic variable and how to extract the image semantic features with semantic rule. Furthermore, it also shows how to manipulate the linguistic values and linguistic expression with the syntax rule. At the end of LEBID processing, each image is associated with a semantic-vector, and each semantic component in