使用下采样帧构成的列向量的PCA系数作为每帧特征（原则上，也可以使用其他类型的关键帧特征），多帧特征构成特征矢量，以高斯概率模型的后验概率衡量相似性、并建立k维树结构的索引进行搜索，文中对噪声、模糊、再压缩、加Logo，及其这些变换的两两组合进行了实验（丢帧很少时也可以），速度很快，适合于大规模因特网视频的搜索。-Use sampling frame column vector of the PCA coefficients as each frame features (in principl

ANN是一个构造数据索引结果的程序，可以实现最近邻、KD树等功能。ANN is designed for data sets that are small enough that the search structure can be stored in main memory (in contrast to approaches from databases that assume that the data resides in secondary storage). Points ar