AI Summary & Key Takeaways
AI InsightThe proposed extraction and searching methods provide robustness to the alterations in video formats and small additions and cuttings in the video content as the typical processing in P2P environment and also achieve uniform distribution and storage load between the peers.
📌 3 Key Takeaways
Abstract
Existing methods in Peer-to-Peer (P2P) networks mainly use file tags or cryptographic hashes of the entire file for video searching and identification. These methods however become insufficient to correctly identify a video when the name and format of the files are changed. In this paper, a distributed solution is proposed for video identification and copy detection in P2P networks, which represents a video file in the network with a set of (64-256) bits, named as perceptual tags. As such information is derived from the perceptual content of the video rather than its bitstream representation as in the case of cryptographic hashes, it provides a robust identification after the alterations in the file names and formats provided that the visual quality of the video is at acceptable levels. The paper first briefly discusses the requirements for a distributed perceptual tagging system considering the low computational power and low bandwidth of internet users. Then, it presents the proposed perceptual tag extraction method using the temporal differences between the video frame averages and the proposed distributed searching scheme for a P2P implementation. The proposed extraction and searching methods provide robustness to the alterations in video formats and small additions and cuttings in the video content as the typical processing in P2P environment and also achieve uniform distribution and storage load between the peers.