利用相似性比较进行视频镜头检测的方法

随着多媒体技术的发展,大量的视频数据涌现。但由于视频数据是非结构性的数据,故而,为了便于视频的检索和浏览,必须对视频进行分段,将镜头作为分段的基本单元是比较合适的。通过改变两帧图像的分块相似度的定义对镜头进行突变检测,然后利用小波变换对分块图像进行多尺度分解,以提取图像的边缘特征;最后结合这两种特征,对镜头进行渐变检测,数值试验表明,所利用的方法对各种类型的视频能够很好的检测出其中的突变型镜头和渐变型镜头。

Similarity Based Algorithm to Detect Video Shot

With the development of multimedia technology,the amount of multimedia video data is enlarged at an ex- plosive speed.As videodata isn't structural,we have to detach it in order to index andbrowse video.It is a good method to use shots as the basic units.We improve the similarity between two frames using the blocked gray histo- gram difference as features to detect the break shots.Then a multi-resolution decomposition using Harr wavelet is applied,several statistical features of the blocked frame edge based on wavelet coefficients are obtained and similari- ty comparison on the features is undertaken.These two kinds of similarity are combined to synthetically predict the similarity between two frames and be used to detect the gradual transitions of shots.From the results of the experi- ments,it is shown that our method can examin all kinds of the shot edges greatly.