An efficient privacy protection scheme for data security in video surveillance

The advancement in video surveillance has raised significant concerns about privacy protection. The existing methods focus on identifying the sensitive region and preserving the behavior of the target, however, they ignore the recoverability of private content. In this paper, we propose a novel and efficient privacy protection scheme for data security in video surveillance, which jointly addresses several key challenges, including de-identification, behavior preservation, recoverability, and compressibility in one unified system. Our method constructs a public stream and a private residual error stream by blurring the private sensitive region. With our scheme, ordinary users could recognize the behaviors in the public identity-protected video stream for surveillance purpose, while authorized users are able to access the recovered private content (e.g., for law investigations). Moreover, the compressed privacy protected region and residual error could be able to save the costs associated with transmission and storage. The extensive experiments on two standard surveillance datasets and a user study demonstrate the effectiveness of our privacy protection system. (C) 2019 Elsevier Inc. All rights reserved.