Video Encoding Scheme Using Autoencoder with Learning Decoded Image of Previous Frame

Abstract

In recent years, research on image processing using deep learning has been active, and an image coding method using neural networks has been proposed. In this paper, we propose a video coding method using an auto encoder. The encoding performance is improved by learning the decoded image of the previous frame during encoding, and video coding is performed. In this study, we conducted computer experiments using 4K video images to compare and evaluate two methods, one using a new autoencoder when learning decoded images and the other using a learned autoencoder. Compared to coding with a learned model, this method can improve the image quality with each successive frame.