摘要
构建多媒体信息系统首先需要对多媒体数据的内容给出尽量全面、详细的表述 ,而现有的多媒体文档描述接口 MPEG- 7标准在信息描述能力上存在不足。针对这一问题 ,根据多媒体文档内容的抽象层次分类 ,提出了一种合理的层次化多媒体文档描述方法 ,并讨论了多媒体文档描述层次间的映射关系。
The paper of Aiello et al [3] is, to our best knowledge, the only published paper that used three layers——physical, structural, semantic——to study image retrieval. In papers trying to improve description of multimedia document, only semantic or structural layer was utilized. We, in this paper, utilize all three layers to achieve description of multimedia document that is more effective than existing methods. Section 1 gives detailed descriptions of physical, structural and semantic layers respectively. These descriptions are in general the same as existing ones found in the open literature except that: (1) in physical layer, we include meta data such as creation information and copyright; (2) in structural layer, we generalize detailed relationships into abstracted relationships; (3) in semantic layer, we use ontology from different knowledge backgrounds to cognize different semantic information of data. Section 2 uses a set of conceptual objects and the presentational structure among them to define what we call MDU (Multimedia Document Unit), a “media connection” between information and raw data. In order to connect the description between physical layer and structural layer and between structural layer and semantic layer, subsection 2.1 defines two mapping relations respectively. Each layer has a number of MDUs. Even in one MDU, there are many relationships among different objects; we define these relationships by traditional DAG (Directed Acyclic Graph). Subsection 2.2 uses this DAG to achieve fuzzy semantic retrieval in addition to what is usually retrieved. Combining the above mentioned definitions, this paper formalizes the presentation of the multimedia document. We believe that this formalized presentation is better than the presentation achievable by the authoritative MPEG 7. The most important characteristic of our formalized presentation is that it can not only provide the overall and detailed description of the content of multimedia data but also operate it in different layers and gr
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2004年第1期6-10,共5页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金 (6 0 37310 8)
国家教育部博士点基金 (2 0 6 990 1)资助
