Abstract: Video shot detection is the first step of content-based video retrieval，and a novel video shot detection method is proposed based on feature information.At first，the color feature and texture feature is gotten from the video frame by wavelet transform，then the similarity between two frames can be calculated form the mutual information of the color feature and the information entropy of the texture feature.This method is robust to flash and significant motion，and the abrupt cuts and gradual cuts can be detected effectively，experimental results show this method has good performance.

摘要： 视频镜头检测是基于内容视频检索的第一步，提出了一种基于特征信息的视频镜头检测方法，该方法首先利用小波变换提取原始视频帧的颜色特征和纹理特征，计算相邻帧颜色特征的互信息量和纹理特征的信息熵差，然后计算镜头间的相似度，该方法可以有效地检测出突变镜头和渐变镜头，对闪光灯及物体和摄像机运动有较强的鲁棒性，实验结果表明该方法具有良好的性能。