中文摘要：

基于内容的视频拷贝检测,目前最流行的方法是基于词袋模型的关键帧内容匹配方法。由于在空间上丢失了视觉词汇的上下文信息,而在时域中,同样丢失了关键帧时域上下文信息,此类方法的精度受到限制。针对这一问题,通过使用一个上下文模型用于计算视频关键帧的空间上下文信息和时域上下文信息,同时将时空上下文信息量化成二进制编码,并通过海明距离实现快速的时空上下文验证。在TREVID—2009视频集上的实验验证了该算法具有较高的效率与准确性。

英文摘要：

The most popular approach for content-based video copy detection is based on bag-of-visual-words model with invariant local features. Due to the neglect of the spatial context information and the temporal context information,these methods are limited. An algorithm of representing the spatial and temporal context information of key frames quantified into binary codes,and spatial-temporal verification is quickly achieved by Hamming distance. Experiments on TREVID—2009 video set demonstrate the proposed algorithm has high efficiency and accuracy.