演化视角下图像的语义表示

Semantic Representation of Images from Evolutionary Perspective

作为记录信息的一种重要形式,图像包含了丰富的语义信息。随着图像资源的不断累积,有效的图像语义表示成为了提高图像检索效率的关键。文章梳理了关于图像描述与表示的相关研究,归纳出现有研究仍存在对过程性知识的表示相对不足,反方向的概括关联表示相对薄弱这两项问题。因此为探究自然有效且可计算的图像语义表示方式,我们提出以"演化"的视角来剖析图像的语义表示问题。文章构建了基于表示过程的层次描述框架,深入分析了该表示方式的重点与难点,并以具体案例示例说明。基于演化视角的图像语义表示重视表示结果与过程,符合用户解读习惯,为更高层次的语义推理和知识发现累积更丰富且可计算的信息。因此该图像语义表示方法具有一定的实用性,可被应用于辅助专家/用户的图像标注等具体实践领域。

As an important form to record information, image resources contain rich semantic information. With the continuous ac- cumulation of image resources, effective image semantic representation has become the key to improving the efficiency of image re- trieval. The article has sorted out the related research on image description and representation, and summed up the two existing problems t： （ 1 ） The representation of procedural knowledge is relatively inadequate; （2） Generalization is relatively weak. There- fore, in order to explore a natural and computable image semantic representation, we discuss it from an evolutionary perspective. This paper first puts forward a hierarchical description framework based on the representation process, then analyses the problems and difficulties of this representation method, and finally presents an example to explain it. This kind of image semantic emphasizes result and process, which accords with the user＇s habit of interpreting images and accumulates richer information for higher-level se- mantic reasoning and knowledge discovery. Thus it can be applied to arenas like image annotation.