A spatiotemporal digital fingerprint extraction algorithm based on visual perception

Currently, there are a lot of digital video data in the internet, and video content safety is still monitored by artificial systems with less efficiency and precision. Digital video fingerprint technology, as a new kind of authentication method based on feature information, is used in digital multimedia works, and causes more and more attention of researchers and enterprises. This paper proposes a novel fingerprint extraction method to monitor video content based on the visual perception. We extract video slices with visual perception features from the video clip, and analyze their frequency domain characterization by Discrete Wavelet Transform, and then extract these conversion coefficients as fingerprint information. Compared with other monitoring technologies, this proposed method extracts features in a segment video rather than in key frames or stable frames. Thus it largely improves the extraction efficiency, and reduces fingerprints size. The results of simulations indicate that this proposed method has high sensitivity, strong robustness and good distinguish ability.