Video Coding Scheme Using Pre-Processing Degradation and Post-Processing Restoration
Han Wang, Luheng Jia, Haoqiang Ren, Kebin Jia · 2023
Extending coding tools beyond but compatible to video coding standards provides a new path to higher coding efficiency. In this paper, we propose an efficient video coding scheme with pre-processing degradation and post-processing restoration. Proper amount of high frequency information is removed before encoding using edge-preserving filter, while the post-processing module exploiting convolutional neural network (CNN) restores degraded frames due to pre-processing and coding artifacts such as quantization error. Experimental results demonstrate that the proposed video coding scheme achieves coding efficiency gain of up to - 9.6% Bjontegaard delta bit rate (BDBR) compared with H.265/HEVC reference model.