基于小波的视频镜头自适应分割

Adaptive Segmentation of Video Shot Based on Wavelet

随着多媒体技术的发展 ,出现了大量的视频数据 .但由于视频数据是非结构性的 ,因此为了便于视频的检索和浏览 ,必须对视频进行分段 ,而将镜头作为分段的基本单位是比较适合的 .通过仔细分析镜头串接的特点 ,根据小波的特性和小波变换检测奇变点的优点 ,提出了应用小波变换检测视频内容差异度函数奇变点的方法对镜头边界进行检测 .首先取一段视频 ,选取亮度直方图对视频的内容进行描述 ,然后提取视频内容差异度变化曲线 ,再进行多分辨小波分解 ,去噪后找出模极大值点 ,经过跟踪可较准确地找出镜头边界 .实验证明 ,该方法能自适应的对视频进行分割 ,对各种类型的镜头边界有较好的检测效果 .

With the development of multimedia technology, the amount of multimedia video data is enlarged at an explosive speed. As video data isn't structural, we have to detach it in order to index and browse video. It is a good method to use shots as the basic units. The analysis of character of shots sequence. It is not a good method using common threshold based because of the variety content of video and the different shot transition type. A novel wavelet based shot boundary detection approach is proposed to overcome these difficulties, it regards the shot change detection as the singularity position detection of frame frame difference function. This makes the selection of threshold easier and gets a good performance. Selecting a video segment and picking up frame to frame difference curve is the first step. in order to improve the speed of the segmentation, an intensity histogram based content description scheme is adopted, the description is sufficient and efficient. The next step is to analyze frame to frame difference function by wavelet transform. Then we filter the noise and find out the module max. Finally we can get the edge of shots accurately by tracking. It is a great method to detach video data adaptively and examine all kinds of shot edges.