由于基于图像高层语义信息的图像检索与传统的图像特征匹配检索相比，在检索的准确性、结果相关性以及降低误检率等方面具有明显的优势，因此高效的图像检索方法应该充分利用图像蕴涵的高层语义信息。为了利用图像的高层语义信息来进行图像检索，在深入研究图像高层语义的低层特征描述的基础上，提出了图像语义的层次划分，并对每个高层语义层提出了语义抽取和检索算法。实验结果表明，该检索算法可以有效地对图像高层语义信息进行提取，并可作为新型高效图像检索系统的一个模型。

Comparing with the traditional image retrieval based on features matching,the method based on the contents of high-level semantic in image has distinguished advance in accuracy,content-relativity and anti-interference.An effective image retrieval method should make fully use of the high-level semantics contents of images.Based on the deep researching on the low-level features' description of image's high-level semantics,in this paper a new model on image's semantic levels is presented,and the extraction and retrieval algorithm for every level of image's semantic is established.The experimental results show that the algorithms can make effective extraction on image's high-level semantic information,and they can be used as a model on novel and effective image retrieval systems.