经过图像分割后,得到的是若干区域和边界.通常把感兴趣的部分叫目标,其余的部分叫背景,最终目的是要让计算机来识别这些目标.为了有效地识别目标,必须要描述这些目标即提供它们的有用信息和相互关系.本文分区域内部描述、区域边界描述和关系描述三个部分,对现有的各种图像描述方法进行全面的讨论,分析了每种方法的原理、优缺点和适用场合,并对图像描述方法的发展前景做了展望.

Some regions and boundaries can be obtained after image segmentation.Usually, region of interest (ROI) is termed object and others background.The ultimate aim is to make a computer recognize these objects. To do this,description of object is necessary, that is to say,to provide some useful information and interact relationship. Comprehensive analyses of existing image description methods by means of three parts--region interior description, boundary description and relationship description are discussed in this paper.In the end,the future of image description methods is presented.