针对基于内容的视觉媒体检索存在的问题,依据多维空间仿生信息学(multi-dimensional space biomimetic informatics,MDSBI)理论的同源连续性规律(principle of homology continuity,PHC),在高维空间中通过几何方法分析视频及镜头和图像,通过子空间方法解决不同视觉媒体的异构性问题,得出视觉媒体检索的实质性工作是在高维空间研究点和子空间的距离问题,并实现适用于视觉媒体检索的镜头分割算法。
The paper analyzed the problems related to visual media retrieval on the basis of contents. The discussion set forth from the PHC of MDSBI,analyzed videos, shots and images in high-dimensional space,and attempted to examine the heterogeneity between different visual media in the perspective of subspace, thus arriving at the conclusion that visual media retrieval should first of all tackle the issue of analyzing the distance between dots and subspace and came up with an algorithm of shot segmentation for visual media retrieval.