Robust video hash extraction

We propose a perceptual video hash function for broadcast monitoring and database search applications. The method consists of binarized low-frequency components of the 3D-DCT transform of video sequences. A video sequence is first brought to a standard frame size and frame length and then processed to yield the hash. Simulation experiments show that the perceptual hash function is unique for different video content, but that it remains invariant under selected signal processing attacks.