Abstract:

In this paper, an emotion-based image semantic query scheme through color description is proposed, which bridges the semantic gaps among the low-level image features, the middle-level semantic indicators and the high-level emotional semantics. In the proposed scheme, images are first segmented into regions through color clustering. Next, some natural language vocabularies are defined to describe the color semantics. Then, the semantic terms for color description are proposed based on several fuzzy membership functions. Finally, by combining the experience of human being about image content and the color description, a high-level emotion-based image semantic query is accomplished by defining a query language. Experimental results show that the proposed scheme is feasible, effective and extensive for the semantic investigation of images.

Key words: image retrieval, emotion-based image query, semantics, color description, fuzzy membership function