基于分块图像小波相对能量和熵的视频镜头边界检测算法

An Algorithm of Video Shot Detection Based on Partitioning Image Wavelet Entropy

提出了对检测窗内图像帧进行分块并进行小波变换,计算尺度j上分块图像的低频、水平、垂直和对角信号的小波相对能量和小波熵的方法,给出了分块图像小波熵的计算公式.采用尺度j上分块图像低频信号小波熵作为特征向量检测视频镜头是否发生转换,即在检测窗内,帧间低频信号小波熵值只发生一次大突变(与设定的阈值相比)就是剪切转换,若发生多次比较大的变化就是渐变转换.采用检测窗内分块图像在尺度j上的垂直、水平和对角高频信号的小波熵作为另一特征量可以区分渐变镜头的类型:淡进/出(fadein/out)或扫换(wipe).采用大量的高清和标清码流进行实验并与基于概率的方法进行比较.结果证明:本文的视频镜头检测方法具有强的鲁棒性、能检测出镜头转换的类型:剪切、淡进/出和扫换,同时有较高的查准率和查全率.

Shot boundary detection is considered to be the primitives for higher level content analysis, indexing,and classification. A new video shot detection algorithm based on wavelet energy and wavelet entropy of the partitioning image is presented. The proposed detector includes twofold. The solution to shot-boundary detection or distinguish the cut and gradual shot using the low frequency of partitioning image wavelet entropy is provided. The formula of partitioning image wavelet entropy is given out. A solution is given to detect the fade in/out or wipe shot boundary using the high frequency of partitioning image wavelet entropy. Major advantages of the detector are its robust and good performance, while there is also the possibility to detect different types of shot boundaries simultaneously. The experiment results indicate that the new algorithm gives better performance than A. Hanjalic.