摘要
提出了一种新的镜头边界检测算法.该算法借鉴信息论的观点,以互信息量和直方图的结合来衡量帧间相似度.通过一自动选取的阈值选出所有可能的镜头切换点,采用动态窗口技术将切变和渐变统一视为切变检测.通过合理的聚类消除噪音所带来的误检现象,提高了镜头的检出率.仿真实验结果表明该算法具有良好的性能.
Based on the theory of information a new shot boundary detection algorithm is proposed. This method is used to indicate the similarity between two image frames through combining mutual information with histogram. All the possible situation of shot switch is selected through a threshold which is generated automatically. Using dynamic window unifies abrupt cuts and gradual cuts, processing all as abrupt transition. A clustering algorithm is applied to avoid the error detection by the noise. Experimental results show this algorithm has good performance.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2007年第10期1949-1953,共5页
Acta Photonica Sinica
关键词
镜头边界检测
互信息量
阈值选择
动态窗口
Shot boundary detection
Mutual information
Threshold selection
Dynamic window
