基于综合特征分析的视频关键帧检索  

Video Key-frame Retrieval Based on Analysis of Combined Features

作　　者：李岩峰[1] 吴渝[1] 徐世龙[2] 

机构地区：[1]重庆邮电大学人工智能研究所,重庆400065 [2]重庆邮电大学通信学院,重庆400065

出　　处：《电视技术》2008年第7期13-16,共4页Video Engineering

基　　金：新世纪人才计划(NCET)项目;重庆市自然科学基金资助项目(2005BB2063);重庆市教委科学技术项目(050509;060517)

摘　　要：综合利用了图像的颜色、形状和纹理特征,实现了对视频关键帧进行基于内容的检索。首先研究关键帧的选取、特征匹配等问题,再从视频处理的层次化结构的底层分析入手,构建了视频的连续帧图像序列,运用时间自适应检测法对镜头的关键帧进行了选取,建立了关键帧图像数据库。实验结果证明该方法性能良好。By comprehensive using of image color, shape and texture features, the content-based retrieval of video key frames is realized. First, the problems like selection and feature matching of key frames are researched. Then, from the basic analysis of layered structure of video processing, the image sequence of continuous frame in video is constructed. Meanwhile, the key frames of lens are selected by using adaptive time detecting method, and finally the database of image key frame is build. The experi-mental results prove the good performance of the proposed method.

关 键 词：视频检索 关键帧 镜头 特征匹配 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]