摘要
多媒体视频关键帧的提取影响视频的整体效果,为便于用户更加快捷浏览、检索感兴趣视频,提出基于决策树的多媒体视频关键帧实时提取方法.利用信息增益比率确定多媒体视频中的关键帧特征,通过ID3决策树确定关键帧特征选取的准则,将被评价为最高信息增益比率的关键帧点当作最新节点进行子树繁衍,采用聚类算法离散化处理ID3决策树的连续属性值,并将选择的关键帧特征输入优化后的ID3决策树,进行关键帧特征分类,实时提取多媒体视频关键帧.实验结果表明:该方法可有效提取代表性视频关键帧,在各类攻击下特征提取准确率始终高于96.9%,并能够准确提取海量多媒体视频关键帧.
The extraction of multimedia video key frames affects the overall effect of the video.In order to facilitate users to browse and retrieve the video of interest more quickly,a real-time extraction method of multimedia video key frames based on decision tree is proposed.The information gain ratio is used to determine the key frame features in the multimedia video.The ID3 decision tree is used to determine the criteria for the selection of key frame features.The key frame points that are evaluated as the highest information gain ratio are used as the latest nodes for subtree reproduction.The clustering algorithm is used to discretize the continuous attribute values of the ID3 decision tree,and the selected key frame features are input into the optimized ID3 decision tree to classify the key frame features and extract the multimedia video key frames in real time.The experimental results show that this method can effectively extract representative video key frames,and the feature extraction accuracy is always higher than 96.9%under various attacks,which can accurately extract massive multimedia video key frames.
作者
陈少伟
王志固
CHEN Shao-wei;WANG Zhi-gu(Department of Culture and Art Creativity,Zhangzhou City College,Zhangzhou 363000,Fujian,China)
出处
《兰州文理学院学报（自然科学版）》
2022年第5期55-60,共6页
Journal of Lanzhou University of Arts and Science(Natural Sciences)
基金
福建省教育科学“十三五”规划2016年度常规课题(FJJKCG16-476)。
关键词
决策树
多媒体
视频关键帧
实时提取
特征分类
信息增益比率
decision tree
multimedia
video key frame
real-time extraction
feature classification
information gain ratio
