Supervised object-based temporal filtering for enhancement of moving facial images

We are interested in recovering accurate facial images from noisy and blurred videos based on 2D and 3D object modelling. Our strategy for enhancement is to employ warping to bring frames of video into correspondence before frame averaging. We use a 2D face model which consists of a set of landmarks that describe the main facial feature points. Experiments to test the strategy are included in the paper. Our results demonstrate that the supervised object-based temporal filtering strategy is simple and effective for enhancement of faces in videos.