一种基于静帧特征分析的视频检索方法

Video Retrieval Based on Static Frame Feature Analysis

由于视频流中嵌入的广告在制作和表现手法上的多样性,使得目前的探测方法效果并不理想。如何找出广告视频的独有特征是探测的关键所在。本文通过对大量广告在制作规律上的统计分析,提出了静帧密集度(Stat-ic frame density,SFD)的概念,静帧密集度是指视频流中1 min内出现静帧的个数,并利用静帧密集度结合镜头强度分析方法检测视频流中的广告序列,实验结果表明此方法具有较强的实用性。但是本文方法主要对时间连续的视频流中嵌套的广告段进行探测,因此不能分辨独立的短时广告,阈值的选取也是影响其性能的关键因素。

The existing methods cannot effectively detect the embedded advertisement videos because of the variety in the design and expression of advertisement.Therefore,it is the key for the advertisement detection to extract the inherent feature in advertisement videos.In this paper,an approach combining the static frame density（SFD）,which means the number of static frames in one minute,and the shot intense method is proposed to detect the advertisement sequences based on the statistic analysis on lots of advertisement videos.Experimental results show the effectiveness and robustness of the proposed method.However,this method is only useful to detect the advertisement sequences in the continuous videos and is affected by the selection of the threshold.