# A low bit rate partial frame-recurrent video coding technique based on enhanced deformable convolutional networks.
Fu Chen (Hosei Univ Tokyo, Japan) - 2022.11.10_

###### tags: `VAADER` `Seminar`


## Video
{%youtube HFOIjke6Y3o %}

## Abstract
In many digital systems, the transmission bandwidth as well as the storage capacity are usually very limited. This introduces challenges for both video transmission and video storage. One of the efficient solution to this problem is to reduce the video resolution before encoding and then increase it after decoding. This is normally referred to as the video post-processing based strategy, and And super-resolution (SR) is a technique that can be used in it to improve the quality of compressed images. In order to seek lower bit-rates and further obtain high-quality videos, we delves into video post-processing and proposes a new approach to improve the compression quality of Versatile Video Coding (VVC) at low-bitrate in this paper. The peculiarity of our proposal is that we didn’t downsample all frames, but only intermediate frames, and we use two highquality frames in a loop to boost intermediate low-quality frames in the deep learned SR network. We obtained an average 61.31% BD-rate reduction on the KristenAndSara sequences compared to VVC, and the average peak signal-to-noise ratio (PSNR) improvement against VVC reached 2.128 dB at low bitrates.