Towards Real-World Face De-Identification
Ralph Gross, Latanya Sweeney · 2007
A wide range of technological advances have helped to make extensive image and video acquisition close to effortless. As a consequence many applications which capture image data of people for either immediate inspection or storage and subsequent sharing have become possible. Along with these improved recording capabilities, however, come concerns about the privacy of people visible in the scene. While algorithms have been proposed to de-identify images, currently available methods are still lacking. In this paper we propose a general framework for the de-identification of images which subsumes a number of previously introduced approaches. Unlike the ad-hoc methods currently used in the field our algorithms aim at providing privacy guarantees. In experiments on illumination-and expression-variant face datasets we show that the proposed algorithms achieve the desired privacy protection while minimally distorting the data.