数字档案馆中视频档案检索框架构建及实现

数字档案馆建设的不断推进,视频档案的数量在急剧增加,视频档案作为档案资料重要组成部分——视频档案的检索成为了数字档案馆建设亟须解决的问题。针对该问题提出了一种基于Contourlet变换的视频检索框架。该方法首先提取视频关键帧,然后对视频关键帧进行3层Contourlet分解,对各高频方向子带求均值和标准差,生成关键帧的特征向量,再计算所有关键帧特征向量的均值生成视频特征,最后利用欧式距离来度量视频的相似度。实验结果证明该方法比小波变换具有更好的综合性能,对于加边框、添加图标、添加字幕、改变比率、镜像、丢帧和噪声等攻击具有较好的鲁棒性。

The Structure and Implementation of Video Archives Retrieval Framework in Digital Archives

As more and more video archives are collected in digital archives,how to retrieve the video that user needs from the massive video archives quickly and accurately has become a problem to be solved in digital archives. Considering this question,a novel video retrieval framework based on Contourlet transform is proposed. First,we extracted key frame from the video,then decomposed video key frames by Contourlet transform and constructed feature vectors by calculating the mean and variance of high-frequency sub-bands. After that,we constructed the video fingerprint by calculating the mean of all key frames feature vectors. Finally,the similarities of videos were computed by the Euclidean’s distance. The experiments show that this scheme is robust to edging,logo insertion,text insertion,ratio,mirroring,drop frames and noise.