Motion-Centric Temporal Frame Interpolation

alt text 

Overview

Our work on highly scalable video coding has lead to some interesting work in temporal frame interpolation (TFI). In contrast to existing TFI methods, we only use motion information in the frame interpolation process. This is achieved by employing state-of-the-art optical flow fields, together with a disciplined use of motion discontinuity information for resolving double mappings as well as to handle disoccluded regions, which are not visible in both reference frames.

We developed different incarnations of the motion-centric TFI approach, which differ in the motion discontinuity representation, as well as how the motion fields used in the prediction of the target frame are constructed.

Related Publications and Links to Project Pages

Journal:

Teaser Image 

D. Ruefenacht, R. Mathew, and D. Taubman, ‘‘Occlusion-Aware Temporal Frame Interpolation in a Highly Scalable Video Setting,’’ APSIPA Transactions on Signal and Information Processing (ATSIP), 2016.
Preprint | Project Webpage

Conference:

Teaser Image 

D. Ruefenacht, R. Mathew, and D. Taubman, “Higher-Order Motion Models for Temporal Frame Interpolation with Applications to Video Coding,” Picture Coding Symposium (PCS), Nuremberg, Germany, 2016.
Preprint | Project Webpage

Teaser Image 

D. Ruefenacht and D. Taubman, “Temporally Consistent High Frame-Rate Upsampling with Motion Sparsification,” IEEE International Workshop on Multimedia Signal Processing (MMSP), Montreal, Canada, 2016.
Preprint | Poster

Teaser Image 

D. Ruefenacht, R. Mathew, and D. Taubman, “Bidirectional, Occlusion-Aware Temporal Frame Interpolation in a Highly Scalable Video Setting,” Picture Coding Symposium (PCS), Cairns, Australia, 2015.
Full text | Presentation Slides | Poster | Project Webpage