镜头分割是基于内容的视频检索的关键步骤,它会直接影响到视频检索的效果。文中介绍了几种常用的镜头分割方法,并且根据YUV颜色空间的各分量可分离的特点,提出了一种改进的基于YUV颜色空间的自适应阈值镜头分割方法。在突变镜头检测模块,为了消除由噪声带来的误检,加入了帧差值比法。在渐变镜头检测模块,使用了滑动窗口值方法。镜头分割的难点是渐变检测,算法在渐变检测上也取得了不错的成果。经过大量实验结果表明,改进的算法对镜头分割有很好的实验效果,算法计算复杂度低,易于实现。
Video shot segmentation is the key procedure of content-based video retrieval (CBVR) ,it affects the results of video retrieval directly. It introduces some commonly used shot segmentation algorithms,presents a shot segmentation approach based on YUV color space while each component of YUV color space can separate. The ratio of interval frames is used to eliminate some misdetection brought by noise. The continuous window value method is used to detect the gradual change happen in the gradual change detection module. The difficult of shot segmentation is the gradual change, the approach has good results on it. As proved by the experimental results, the ap- proach is brief and effective for shot boundary detection.