摘 要:如何从动态视频图像大量的冗余信息中提取关键信息,进而有效地存储和检索视频成为当前本领域研究的热点。模型通过对图像帧进行分块,再对块图像帧进行预处理,基于熵密度的比较选择初始聚类中心,进而确定初始聚类半径,应用归一运算 , 合并同类图像帧以生成关键帧,用来表征图像的主要内容。实验表明该文模型能够减少动态视频信息中的冗余度,同时还能有效地还原视频真实内容,对于视频存储和检索具有非常重要的意义。
关键词:K-Means;图像熵;关键帧;视频检索
DOI:10.19850/j.cnki.2096-4706.2021.03.003
基金项目:武汉省科技局科技项目(201901 0702011291);广东省自然科学基金项目(2019A15 15010411)
中图分类号:TP391.41 文献标识码:A 文章编号:2096-4706(2021)03-0009-05
Key Frame Extraction Model of Dynamic Video Based on Improved K-Means
XIANG Dong1 ,JI Jing1 ,ZHANG Jingrui 2 ,OUYANG Quan1
(1.Wuhan Xingtu Xinke Electronics Co.,Ltd,Wuhan 430073,China; 2.School of Aerospace Engineering,Xiamen University,Xiamen 361102,China)
Abstract:How to extract the key information from large amount of redundant information of dynamic video images,and then effectively store and retrieve video has become research hotspots in the field at present. The model divides image frames into blocks,and then pre-process block image frames. Based on the comparison of entropy density,the initial clustering center is selected,and then the initial clustering radius is determined. The normalization operation is applied and the same kind of image frames are combined to generate key frames,which are used to represent the main content of the image. The experiment shows that the model proposed in this paper can reduce the redundancy of dynamic video information and effectively restore the true content of video. It is very important for video storage and retrieval.
Keywords:K-Means;image entropy;key frame;video retrieval
参考文献:
[1] NAGASAKA A,TANAKA Y. Automatic video Indexing and Full video Search for Object Appearances [C]// Visual Database System,Elsevier,1991:113-127.
[2] RACHIDA H,ELBOUSHAKI A,AFDEL K,et al. An efficient method for video shot boundary detection and keyframe extraction using SIFT-point distribution histogram [J]. International Journal of Multimedia Information Retrieval,2016,5(2):89-104.
[3] BARBIER T,GOULARTE R. KS-SIFT:a keyframe extraction method based on local features [C].Proc of IEEE International Symposium on Multimedi. Washingt on DC:IEEE Computer Society,2015:13-17.
[4] CHEN W,FANG M,LIU Y,et al,Monocular Semantic SLAM in Dynamic Street Scene Based on Multiple Object Tracking [C]//2017 IEEE International Conference on Cybernetics and Intelligent Systems(CIS)and IEEE Conferenceon Robotics,Automation and Mechatronics,Ningbo,2017:599 - 604.
[5] 王俊玲,卢新明.基于语义相关的视频关键帧提取算法 [J]. 计算机工程与应用,2021,57(4):192-198.
[6] 田丽华,张咪,李晨 . 基于运动目标特征的关键帧提取算法 [J]. 计算机应用研究,2019,36 (10):3183-3186.
[7] 卢小平,卢遥,焦金龙等 . 基于重叠区域相关系数的视频 影像关键帧提取算法 [J]. 武汉大学学报 (信息科学版),2019, 44(2),260-267.
[8] 白慧茹,吕进来.基于聚类方法改进的关键帧提取算法 [J]. 计算机工程与设计,2017,38(7):1929-1933.
[9] 蒋元友 . 一种基于聚类的关键帧提取算法 [J]. 数字技术与 应用,2014(11):126-127.
[10] MACQUEEN J.Some methods for classification and analysis of multivariate observations [J].Berkeley Symposium on Mathematical Statistics and Probability,1967:281-297.
作者简介:向东(1983—),男,汉族,湖北恩施人,高级工 程师,博士,主要研究方向:人工智能、计算机应用、信息化系统。