基于镜头的鲁棒视频广告检测

随着多媒体技术的发展，自动检测出数字视频节目里面嵌入的广告是很具挑战性的研究．然而，由于嵌入的广告的制作方式和表现手法的多样性，很多自动检测模型的实验结果往往不甚理想．为了提高检测系统的鲁棒性，提出了3阶段广告检测系统．首先，提出了基于区域特征重要性的镜头检测算法（RBFID，region-based feature importance detection），实现视频播放中突变镜头和消隐镜头的检测，同时从每个镜头提取出一些统计特征用来标识镜头．然后，利用SVM的优异分类特性实现镜头分类．最后为了能得到精确的广告视频段，利用广告视频在内容和时间上的连续性来消除错分的镜头，然后将广告镜头整合成广告视频段．本系统在30个电视节目的片段上进行验证，实验结果表明此广告检测系统具有实用性．

Video commercial detection based on the robustness of shot

Automatic detection of commercials embedded in digital video materials is a challenging task with the development of retrieval of multimedia data. However, because of the diversity of production modes and expression methods of commercials, the performances of current detection systems are inadequate. In this paper, a three-phase system for commercial detection is suggested to improve robustness of the detec- tion system. Firstly, two Region-based Feature Importance Detection schemes are proposed to detect cut shots and dissolved shots respectively and some statistical features are also extracted to mark shots. Sec- ondly, an SVM classifier is applied to classify these shots. Finally, in order to obtain commercial segments more accurately, a statistical comparison of time and content of commercials is used to eliminate falsely cut shots. Finally, these commercial shots are integrated into commercial video segment. Test results on 30 TV video segments show the effectiveness of the suggested system.