基于最邻近帧质量增强的视频编码参考帧列表优化算法  

Reference frame list optimization algorithm in video coding by quality enhancement of the nearest picture

作　　者：霍俊彦[1] 邱瑞鹏 马彦卓[1] 杨付正[1] HUO Junyan;QIU Ruipeng;MA Yanzhuo;YANG Fuzheng(State Key Laboratory of Integrated Services Network,Xidian University,Xi’an 710071,China)

机构地区：[1]西安电子科技大学ISN国家重点实验室,陕西西安710071

出　　处：《通信学报》2022年第11期136-147,共12页Journal on Communications

基　　金：国家自然科学基金资助项目(No.62101409,No.62171353)。

摘　　要：帧间预测是视频编码的核心模块,其利用参考帧的重建样本来预测当前图像样本,从而通过传输少量预测残差数据表示复杂视频内容。在有损视频编码中,参考帧质量受到量化失真的影响,导致预测精度较差,影响编码性能。针对低时延视频业务,提出一种基于最邻近帧质量增强的参考帧列表优化算法,通过基于深度学习的卷积神经网络增强与当前帧最邻近参考帧的质量,并将增强后的高质量帧整合到当前帧的参考帧列表中,提高了帧间预测精度。以高效视频编码H.265/HEVC参考软件平台HM16.22为参考基准,所提算法在Y、Cb、Cr这3个分量上可分别节省9.06%、14.92%、13.19%的编码码率。Interframe prediction is a key module in video coding,which uses the samples in the reference frames to pre-dict those in the current picture,thus helps to represent the complex video by transmitting a small amount of the predic-tion residual.In lossy video coding,the qualities of reference frames are affected by the quantization distortion,which lead to poor prediction accuracy and performance degradation.Targeted at the low latency video services,a reference frame list optimization algorithm was proposed,which enhanced the quality of the nearest reference frame by a deep learning-based convolutional neural network,and integrated the enhanced reference frame into the reference frame list to improve the accuracy of interframe prediction.Compared with H.265/HEVC reference software HM16.22,the proposed algorithm provides BD-rate savings of 9.06%,14.92%and 13.19%for Y,Cb and Cr components,respectively.

关 键 词：H.265/HEVC 深度学习 帧间预测 参考帧列表 

分 类 号：TN911.7[电子电信—通信与信息系统]