Abstract:

As an important form to record information, image resources contain rich semantic information. With the continuous accumulation of image resources, effective image semantic representation has become the key to improving the efficiency of image retrieval. The article has sorted out the related research on image description and representation, and summed up the two existing problems t: (1) The representation of procedural knowledge is relatively inadequate; (2) Generalization is relatively weak. Therefore, in order to explore a natural and computable image semantic representation, we discuss it from an evolutionary perspective. This paper first puts forward a hierarchical description framework based on the representation process, then analyses the problems and difficulties of this representation method, and finally presents an example to explain it. This kind of image semantic emphasizes result and process, which accords with the user’s habit of interpreting images and accumulates richer information for higherlevel semantic reasoning and knowledge discovery. Thus it can be applied to arenas like image annotation.

Key words: Image semantic representation, Semantic evolution, Hierarchical description framework, Image annotation, Semantic data model