The Methods of Semantics Processing in Content-Based Image Retrieval
基于内容的图象检索中的语义处理方法

The goal of content-based image retrieval(CBIR) is uppermost to reduce the semantic gap between the simple visual features and the abundant semantics delivered by an image, and a critical point in the advancement of content-based retrieval is the image semantic modeling and extraction. This paper reviews the state of the art of image retrieval using semantics mainly focusing upon two aspects: image semantic modeling and image semantic extraction. In the paper, image semantic model is generalized firstly as three main components: image semantic knowledge, image semantic hierarchical model and semantic extraction hierarchical model. Then some typical methods of semantics extraction of are analyzed in detail by classifying into five classes: semantic by user interaction, user query as visual semantic template, objects and their layout recognition, scene and event semantics extraction, emotion semantics extraction, and some limitation of them are pointed out. Finally, three critical problems including object modeling and recognition, semantic knowledge base and user retrieval model faced in image semantic processing are explained, and some resolved strategies are presented elementarily.