基于窗帧差的镜头边界系数模型及其应用  

Shot Boundary Coefficient Model and Its Application Based on Windowed Frame Differences

作　　者：方勇[1] 戚飞虎[1] 冉鑫[1] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200030

出　　处：《电子学报》2006年第5期810-816,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60072029;60271033)

摘　　要：针对视频结构分析提出了一种新的镜头边界系数模型.首先,按照特定的计算方法计算当前帧的窗帧差,使得窗帧差在镜头内与镜头边界呈现不同的特征;然后,采用相应的镜头边界检测算子计算镜头边界相似系数,镜头边界相似系数越大,则该帧越可能是镜头边界;最后,为了提高镜头边界相似系数的检测能力,根据镜头边界相似系数定义了镜头边界系数.镜头边界系数具有良好的特性,可以单独检测镜头边界,也可以与传统的镜头边界检测方法相结合,为镜头边界检测提供了一种鲁棒的框架.实验结果表明,基于镜头边界系数模型的镜头边界检测方法能明显改善镜头边界检测结果.To aim at video structure analysis, a new shot boundary coefficient model is proposed. Firstly, windowed frame differences are calculated using the special approaches,which have different characteristics for intra-shot frames and shot transition frames respectively. Secondly, shot boundary detection operators are used to calculate shot boundary similarity coefficients. If shot boundary similarity coefficients are bigger, corresponding frames are more probably shot boundaries. Thirdly, to improve shot boundary similarity coefficients, shot boundary coefficients are defined based on shot boundary similarity coefficients. Shot boundary coefficients have many good characteristics, and are used to not only detect shot boundaries independently, but also combine traditional shot boundary detection methods for shot boundary detection. Therefore, a robust shot boundary detection framework is provided. Experimental results show that shot boundary detection methods, based on the shot boundary coefficient model, can improve the performance of shot boundary detection.

关 键 词：视频分割 镜头边界检测 镜头边界系数 视频检索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]