Super-resolution Model-based Versatile Video Coding for Light Field Video

In this paper, we propose a novel light field (LF) super-resolution model for a downsampling-based LF video coding framework by introducing a feature selection and shuffling mechanism based on our super-resolution model. It is shown to achieve a good performance on the downsampling-based LF video coding framework benefiting from its ability for learning of the most useful features and effective removal of the artifact caused by the VVC encoding and downsampling process. The proposed scheme is shown to achieve BD-rate savings of about 39% and 28% on AI-Main and RA-Main conditions for LF video coding against the anchor of VVC.