一种改进的视频关键帧提取算法研究  

Study on an Improved Algorithm of Video Keyframe Extraction

作　　者：瞿中[1,2] 高腾飞[1] 张庆庆[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]重庆邮电大学移通学院,重庆401520

出　　处：《计算机科学》2012年第8期300-303,共4页Computer Science

基　　金：重庆市科委自然科学基金计划项目(CSTC;2010BB2399);重庆市教委科学技术研究项目(KJ110528)资助

摘　　要：视频镜头分割和关键帧提取是基于内容的视频检索的核心问题。提出了一种改进的关键帧提取算法,其为视频检索奠定了基础,镜头分割部分采用改进直方图方法及基于像素方法的综合方法。首先,通过结合直方图交集及非均匀分块加权的改进直方图方法,根据视频内容将视频分割为镜头;然后,利用基于像素的帧差法,对得到的检测镜头进行二次检测,优化检测结果;最后,在HSV颜色空间的基础上,计算每个镜头内每帧的图像熵,从而确定关键帧序列。实验结果表明,提出的改进算法所得到的关键帧结构紧凑且分布均匀。Video segmentation and keyframe extraction are the core issues of Content-based Video Retrieval（CBVR）.We proposed an improved approach of key-frame extraction for video retrieval,adopting the synthesis method based on histogram-based method and pixel-based method in video shot segmentation.In the method,videos are firstly segmented into shots according to video content,by our improved histogram-based method,with the use of histogram intersection and nonuniform partitioning and weighting.Then,the obtained results are secondly detected by pixel-based method to optimize the results.At last,within each shot,key-frames are determined with the calculation of image entropy of every frame in HSV color space.Our simulation results prove that the extracted key frames with our method are compact and uniformly distributed.

关 键 词：视频镜头检测 关键帧提取 帧差法 图像熵 视频检索 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]