A Facial Privacy Protection Framework Based on Component Difference and Template Morphing

Abstract

Aiming to countermeasure facial privacy disclosure of the shared images in social media, a face privacy protection framework based on component difference and template morphing is proposed. For a shared facial image that requires privacy protection, its facial attributes are first detected, and then the most suitable face template is searched from a pre-built facial image template library. After that, the key points of the facial image and the face template are detected, and they are implemented for facial components segmentation. Finally, the facial components of two images are morphed according to the privacy protection level and the optimal morphing sequence determined by the component difference. Experiments and analysis are performed to an implementation of the framework. The results show that it can effectively protect the facial privacy meanwhile keep the visual quality of the image. It has great potential to be applied for privacy protection of the shared facial images in social media.