一种基于内容视频检索中的镜头分割算法

A Shot Boundary Detection Algorithm in Content-Based Video Retrieval

提出了一种统一的可以检测出切变、渐变的视频镜头检测算法。首先,通过颜色局部累积直方图与纹理相结合的方法获得视频的帧间差,对于切变检测,通过已经计算得到的帧间差为基础,采用双重窗口检测算法实现,对于渐变的检测,采取了统一的双阈值检测方法对所有渐变类型进行了检测。通过多次试验,证明了用本文算法进行视频镜头检测能取得比较好的效果。

This paper presents a unified shot boundary detected algorithm. At first, the difference of neighboring frames is gotten by using the combination of the colorful local cumulative histogram and texture information. Then for the abrupt change detection and based on the calculated difference between frames, the paper uses the dual-window detection algorithm, for gradual change detection, to take a unified dual-thresbold detection method for all types of gradual change detection. Through repeated tests, it proves that using the above algorithm can achieve good results of the shot boundary detection.