Protecting the privacy of humans in video sequences using a computer vision-based de-identification pipeline

作者：

Highlights：

• A computer-vision based pipeline that de-identifies humans in videos is proposed

• Precise localization of humans is performed using an improved GrabCut algorithm

• We alter the appearance of humans while preserving the naturalness of the scene

• De-identification using neural art conceals soft and non-biometric identifiers

• The proposed pipeline thwarts identification both by humans and by machines.

摘要

•A computer-vision based pipeline that de-identifies humans in videos is proposed•Precise localization of humans is performed using an improved GrabCut algorithm•We alter the appearance of humans while preserving the naturalness of the scene•De-identification using neural art conceals soft and non-biometric identifiers•The proposed pipeline thwarts identification both by humans and by machines.

论文关键词：Privacy protection,De-identification,Computer vision,Video processing

论文评审过程：Received 12 October 2016, Revised 5 May 2017, Accepted 27 May 2017, Available online 17 July 2017, Version of Record 17 July 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.05.067