Facial Video Frame Interpolation Combining Symmetric and Asymmetric Motions

We propose a facial video frame interpolation algorithm, which combines symmetric and asymmetric motions to complement each other. We generate multi-scale contextual features and intermediate frame candidates using both kinds of motion vectors. Then, through a frame synthesis block, we refine those intermediate frame candidates, by exploiting the contextual features, to produce the final output. Extensive experimental results demonstrate that the proposed algorithm provides state-of-the-art performance by combining both symmetric and asymmetric motion information effectively.