Synthetic and privacy-preserving visualization of video sensor network outputs

We propose a complete framework for people tracking in video surveillance networks that preserves privacy by mapping people recorded by a video sensor network in the map of the corresponding surveilled areas. Thanks to this approach, it is possible (1) to synthesize multiple video outputs into a unique picture, and (2) to control privacy by offering the possibility to filter information to viewers. A set of physical attributes, namely soft biometric characteristics, is used to provide the system with tracking capabilities. We demonstrate the feasibility of our approach by showing a real case application scenario.