Towards a comprehensive survey of the semantic gap in visual image retrieval

Abstract

This paper adopts the premise that the 'semantic gap' is an incompletely surveyed feature in the landscape of visual image retrieval, and proposes a framework within which this deficiency might be made good. Simple classifications of types of image and of types of user are proposed. Consideration is then given in outline to how semantic content is realised by each class of user within each class of image. The argument is advanced that this realisation finds expression in perceptual, generic interpretive and specific interpretive content. This analytic framework provides the basis for the specification of a broadly encompassing evaluation study, which will employ the image/user type classification and the expert domain knowledge of selected user groups in the construction of segmented test collections of real queries, images and relevance judgements. From this study should come a better-informed view on the nature of semantic information need, and on the representation and recovery of semantic content across a broad spectrum of image retrieval activity.