提出了一种基于关键帧颜色和纹理特征的视频拷贝检测方法。首先通过子片段方法提取视频的关键帧,然后将关键帧分成3个子块,提取每个子块的三维量化颜色直方图,通过直方图相交法来进行颜色特征的匹配。对检索得到的结果视频关键帧进行纹理特征提取,通过其灰度共生矩阵的角二阶矩和熵来表征其纹理特征,纹理特征的匹配可进一步过滤不相关的视频。实验结果表明,该方法效果好、稳健性强且可应用于多种类型的视频。
A video copy detection method based on color and texture features of key frames is proposed. Firstly, key frames are extracted based on clipmethod. Then, key frames are divided into 3 blocks and three-dimensional quantized color histograms are extracted from the blocks. Color matching is based on histogram intersection. Texture features are further extracted from the key frames of the resulting videos and are represented with angular second moment and entropy of the co-occurrence matrix. With the matching of texture features, more irrelative videos are filtered. Experiments show that this method is effective, high robust and can be used for various types of videos.