基于H.266/VVC的多帧质量补偿编码方法  

Multi-frame Quality Compensation Coding Method Basedon H.266/VVC

作　　者：田双星 丁丹丹[1] TIAN Shuangxing;DING Dandan(School of Information Science and Technology,Hangzhou Normal University,Hangzhou 311121,China)

机构地区：[1]杭州师范大学信息科学与技术学院,浙江杭州311121

出　　处：《杭州师范大学学报（自然科学版）》2024年第5期472-479,共8页Journal of Hangzhou Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(62171174).

摘　　要：文章提出了一种基于H.266/VVC标准的多帧质量补偿编码方法,旨在提高视频编码效率并优化重建帧质量.通过引入光流法提取前后帧的时间信息合成虚拟参考帧,结合Transformer模块与原始帧引导模块,有效融合了合成的虚拟帧与重建帧,并提取了原始帧上的高质量信息,以指导生成更高质量的重建帧.在H.266/VVC官方测试集上的训练和测试结果表明,所提出的方法能够获得显著的性能增益.This study introduced a multi-frame quality compensation coding method based on the H.266/VVC standard,aiming to enhance video coding efficiency and optimize the quality of reconstructed frames.By incorporating optical flow to extract temporal information from preceding and succeeding frames to synthesize virtual reference frames,and integrating a Transformer module with an original frame guidance module,this method effectively fused the synthesized virtual frames with reconstructed frames and extracted high-quality information from the original frames to guide the generation of higher quality reconstructed frames.Training and testing on the official H.266/VVC test set demonstrated that the proposed method achieved significant performance gains.

关 键 词：视频编码 质量补偿 卷积神经网络 原始帧引导 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]