Abstract: Semantic gap has become a bottleneck of content-based image retrieval.In order to bridge the gap and improve retrieval accuracy,a map from lower-level visual features to high-level semantics should be formulated.This paper provides a comprehensive survey on semantic mapping.Firstly,an image retrieval framework integrated with high-level semantics is presented.Secondly,image semantic description is introduced in two aspects:image content level-models and semantic representations.Thirdly,as the emphasis of this paper,semantic mapping approaches and techniques are investigated by classifying them into four main categories in terms of their characteristics.Various ideas and models proposed in these approaches are analyzed.In addition,advantages and limitations of each category are discussed.Finally,based on the state-of-the-art technology and the demand from real-world applications,several important issues related to semantic image retrieval are identified and some promising research directions are suggested.