Abstract: The research focus of CBIR is how to understand the content of image，digest the features of image and organize those features to retrieve.The result of retrieval is often discarded，which can not be used for the future retrieval of the similar images.This paper proposes a novel scheme for semantics saving of the results of the image retrieval，which uses the similar images of the retrieval results to form the complex network whose nodes are the regions of the images and uses the community finding algorithm to get the semantics communities to save the semantics concept of the retrieval results for matching the similar image of the same semantics in the future retrieval.It has shown from the experiments that the complex networks have the character of small world and can quicken the retrieval process for retrieving the similar semantics images.

Key words: content-based image retrieval, complex networks, small world network, semantics saving

摘要： 目前多数基于内容的图像检索研究是在如何理解图像的内容，挖掘图像内容的特征，并组织这些特征用于图像检索上。检索得到的结果往往丢弃不顾，不能用于未来相似语义图像的检索。提出了一种新颖的基于语义保存的图像检索方案：将检索得到的多个相似图像组成相似图像网络，并运用复杂性网络的子网络分割方法，划分出语义子网络，找出语义概念并保存；检索未来相似内容的图像只需匹配保存的图像语义概念。实验表明，检索后得到的图像网络具有小世界网络的特征；保存的图像语义能准确地匹配相似语义内容图像，并能极大地加快检索相似语义图像。

关键词: 基于内容图像检索, 复杂性网络, 小世界网络, 语义保存