基于深度学习的AVS3视频编码技术研究

Research on AVS3 Video Coding Technology Based on Deep Learning

随着超高清视频的发展和5G时代的到来,海量数据的存储和传输对视频编码技术提出了更高的要求。近年来,深度学习在视频压缩编码中的应用逐渐成为新的研究方向,为视频编码领域的发展注入了活力。借助新一代视频编码标准AVS3研究的契机,文章介绍了AVS3中的深度学习相关技术,包括帧内预测、帧间预测和环路滤波方面,并对AVS3参考编码软件中的深度学习环路滤波工具进行了性能测试。

With the development of UHD video and the arrival of 5G era,numerous video data brings great challenge for data storage and network transmission.In recent years,video coding method based on deep learning is becoming a new research direction gradually,which has injected vitality into the development of video coding.While a new generation of video coding standard AVS3 is established,this paper introduces key technologies based on deep learning in AVS3,including intra prediction,inter prediction and loop filtering.This paper also tests the performance of loop filter tool based on deep learning in AVS3 reference coding software.