一种自适应的镜头探测系统的设计与实现

镜头探测是视频内容结构化分析的第一步,也是非常关键的一步。目前的镜头探测方法根据处理方式不同,可分为压缩域和解压域两种,而研究的重点多集中在mpg格式的视频文件上。随着压缩标准的不断成熟完善,出现了越来越多的视频压缩格式,如avi、wmv、asf等等。对于这类格式的视频文件如何进行视频内容结构化分析成为了迫切需要解决的一个问题。该文设计并实现了一种能适应多种视频文件格式的镜头探测系统,该系统采用先解压、后探测的方案。解压通过微软提供的DIRECTXSDK进行,接下来进行的镜头探测在解压后的像素域上进行。实验证明,该方法是行之有效的,它为多格式的视频流分析提供了一种解决思路。

Design and Realization of an Adaptive Shot Detection System

Shot detection is the first and the fundamental step in video content and structure analyses.According to different viewpoint of process,nowadays shot detection methods are mainly based on compressed domain and decompression domain,and emphasize mostly on the format of mpg stream.While along with the development of MPEG standard,more and more video compress formats appear,such as avi,wmv ,asf,etc.How to analyze these types of video files becomes an emergent problem.In this paper,a shot detection system that is adaptive to various video formats is designed and realized.It adopts the idea of decompressing first,detecting later.Decompressing is carried through DIRECTX SDK,which is provided by Microsoft Corporation.And then detecting shot boundaries in pixel domain.Experiments prove that this method is effective and robust;it provides a solution for various formats of video analyses.