中文摘要：

为了解决空域特征与时域特征只能片面地表征视频内容的问题，提出了一种基于时空代表帧序列的视频近拷贝检测方法．首先，根据帧间颜色卡方差与灰度OM（ordinalmeasurement）差别进行场景分割；其次，对分割后的视频进行降帧率操作，对于每个分割后的场景，进行分段合成，并根据加权的灰度互信息量进行筛选，得到每个场景的时空代表帧（representativeimage），并对视频代表帧提取灰度的OM、均值、标准差、对比度以及颜色分布描述子和边缘OM序列等特征；最后，对视频的特征进行对比，根据特征的相似程度来判断检测视频是否为近拷贝视频．实验结果表明，该方法具有较好的鲁棒性和检测能力．

英文摘要：

To solve the problem that temporal or spatial features only characterized the video content in an unilateral way, a method of near-duplicate video detection based on representative-sequence was presented. Firstly, the video was divided into different scenes according to the variance of the Chi- squared color histogram and grayscale OM （ordinal measurement）, and then the frame rate was decreased to a particular rate. For each scene, every certain number of frames were merged into one TIRI frame, and then the weighted grayscale mutual information was calculated to select the representative frames from the TIRI frames. Next, several features were extracted from the representative frames for matching, including the mean, standard deviation, contrast, and an OM sequence of grayscale, as well as color distribution descriptors, and an OM sequence of edge information. Finally, whether or not the video was a near-duplicate was determined by the match degree of the characteristics. Results show that the algorithm presented possesses good robustness and detection capabilities.