A robust and low-complexity video fingerprint for multimedia security

摘要

Multimedia security means protecting multimedia information and information systems from unauthorized access, use, disclosure, disruption, modification, inspection, recording and destruction. The “Video Copy Detection” system is based on detecting video copies from a video sample to avoid copyright violations. Video fingerprinting is one of methods for video copy detection, including steps of video fingerprinting extraction and matching. Unlike previous video fingerprinting methods, a novel efficient and robust video fingerprint based on region of interest (ROI) and clustering is presented in this paper. In video fingerprinting extraction step, ROI is extended as basic unit for extracting video fingerprints. Features of Freak (fast retina keypoint) extracted from ROIs and clustered by spectral clustering later will be used as video fingerprints. In the matching step, we extend the method based on the classical cosine similarity with a significant change to measure the distance between video fingerprints. Results show that the proposed video fingerprinting exceeds several representative methods on its low-complexity and its robust on perceived similar alterations especially on text insertion, logo insertion, cropping, padding, frame dropping and fps changing et al.