Privacy Protection

Privacy concerns on life logs involve all sorts of collected data. Let us discuss all such issues and technical solutions on this page

Image data

For images taken by ViconRevue or iPhone, let us focus on the people inside rather than objects as Google does on the street view image data like below. Google_StreetView_Privacy_Protection_Example.png

Google detects faces from the scene and blurs them to protect the privacy. We may do the similar approach but the things to consider compared with Google is that photos have boundaries. Googles images are seamless in a sense that it capture the entire 360 degree views in a continuous manner which means that if a people in there, his entire presence would be captured by the system, whereas the photo taken by our system has a limited dimensions of both width and height. Thus the performance of any face detection or face feature detection algorithm would suffer such incomplete image features.

So we leveraged this problem using multiple face feature detectors including frontal face detector, left eye detector, right eye detector, nose detector, lip detector and upper body detector. See the below video processing example.


Algorithm evaluation

{{#widget:Google Spreadsheet |key=0ApMuCRL-y7qxdG5hNHpORXhkMXhVNjN1X2VXaC1seEE |width=700 |height=450 }}