电视节目自动分割算法

An Automatic Program Segmentation Method for TV Video

近年来 ,视频内容的结构化分析逐渐成为一个研究热点 传统意义上认为场景是最高层次的视频处理单元 然而随着研究工作的不断深入 ,需要提出一种更高层次的视频处理单元———节目 ( program) 通过节目管理大量存在的场景 ,形成更丰富的层次 ;另外 ,节目还可以作为视频节目分类算法的输入 ,从而实现从分割到分类的全部自动化 在传统层次结构的基础上 ,进一步拓宽视频处理单元的概念 ,在场景层与视频层之间增加节目层 并且提出一种针对特定电视频道的视频流的节目自动分割算法 。

During the recent ten years, analysis of video content has become a hot research topic Generally speaking, in the video hierarchical structure, scene is the highest layer of video processing However, with the current research progress in video processing, It is believed that a higher layer video processing unit, “Program”, is needed to represent a video “program” is defined as one of more continuous scenes For example, in a several hour very comprehensive video stream, there are thousands of scence It is difficult to manage this large amount of information without the program layer Another example is video genre verification Almost all of the proposed algorithms in this research area are based on the assumption that programs have been manually segmented So, implementing automatic program segmentation is of great benefit in reality A new video processing unit, “program”, is proposed to be inserted between scene layer and video layer A TV program specific automatic segmentation algorithm is also proposed