摘要
视频镜头切分及关键帧提取是基于内容的视频检索的基础。通过使用色彩直方图向量来表征视频帧特征,使用余弦夹角来度量视频帧的相似性并建立相邻视频帧的相似性数据序列,再通过小波变换寻找其中的突变数据点或数据点集来完成对视频镜头的切分,最后通过最小直方图特征标准差在每个镜头中提取关键帧。实验表明,该方法能有效地找到镜头的边界,所提取的关键帧可以表达镜头的主要内容。
Video shot segmentation and key frame extraction are the basis of content-based video retrieval.By using histogram vector to represent the video frame characteristics,the cosine angle value is used to represent the similarity of neighbouring frames and establish the similarity series,using the wavelet transformation to find the abrupt data point or data point set,camera shots can be founded in video.At last,the minimum standard deviation is used to extract the key frames in each shot.Experiment indicates that this method is effective and can find the camera shots,also the extracted key frames are effective too.
出处
《计算机与数字工程》
2016年第9期1816-1820,1870,共6页
Computer & Digital Engineering
关键词
小波变换
视频
镜头切分
关键帧
wavelet transformation
video stream
shot segmentation
key frame
