图像内容理解与检索技术

Image Content Understanding and Retrieval

大量图像的出现使得图像检索的需求越来越强烈。图像检索系统经历了基于文字的检索和基于底层特征的检索两个阶段。新的检索系统通过机器学习等技术，综合利用了视觉特征和高层语义概念进行更精确的语义检索。本文从图像底层特征表示与抽取、高层语义特征抽取、相关反馈等方面介绍了当前图像检索系统的基本进展，并对图像检索的发展趋势进行了展望。

Large amount of images urges the requirement for image retrieval. Previous image retrieval systems were based on text description or low features within images. New generation image retrieval systems make use of some new technologies such as machine learning methods, combine visual features with high level concepts to get more accurate results. This paper surveys the research progresses in this field, especially on low level image feature selection and representation, high level semantic feature extraction and feedback. Some research trends are also discussed in the end of the paper.