联合多特征的未来视频快速编码  

Joint multi-feature fast coding for future video coding

作　　者：崔鑫 彭宗举[1] 陈芬[1] CUI Xin;PENG Zong-ju;CHEN Fen(Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo 315211, China)

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《光学精密工程》2019年第4期990-999,共10页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61771269;No.61620106012;No.61671258);浙江省自然科学基金资助项目(No.LY17F010005)

摘　　要：与前一代高效视频编码标准相比,未来视频编码标准增加了新的二叉树加四叉树编码结构。针对该结构虽然显著提升了超高清视频编码效率,但是大幅增加了编码复杂度的问题,提出了联合多特征快速编码算法。该算法联合多个编码特征的后验概率信息,估计当前编码单元的划分方式,通过提前终止若干编码单元的划分来节省编码时间。此外,针对单个特征贝叶斯分类不准确的问题,算法在简化联合概率模型的同时提升了分类算法的准确度。实验结果表明:在随机访问配置、低延迟P配置和低延迟B配置下,算法可以平均减少35.7%,25.6%和26.7%的编码复杂度,而BDBR只分别增长了4.3%,3.1%和2.89%。算法在保证视频主观质量的前提下,节省了编码时间。Future video coding considerably improves compression efficiency over that of the previous high-efficiency video coding standard. The quadtree plus binary tree structure of future video coding is more consistent with the texture of high-definition video but at the cost of tremendous computational complexity. In this study, a joint multi-feature classification algorithm was proposed to reduce the complexity of the quadtree plus binary tree block structure in future video coding. The partition mode of the current coding unit was estimated using the proposed algorithm, and the partition of the coding unit was then terminated. Consequently, coding time was conserved by the proposed algorithm. In general, classification with a single feature was inaccurate. Thus, a joint multi-feature classification model was used in the proposed algorithm. The proposed algorithm not only improveed the accuracy of the classification, but also simplified the joint probability model. Experimental results show that our algorithm can reduce the coding complexity on average by 35.7%, 25.6%, and 26.7% with only a 4.3 %, 3.1%, and 2.89% Bj ntegaard delta bit rate increment with the random access main, low delay P, and low delay B configurations, respectively. The encoding time is conserved by the proposed algorithm based on subjective video quality.

关 键 词：未来视频编码 联合多特征快速编码算法 二叉树加四叉树结构 快速编码 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术] TH691.9[自动化与计算机技术—计算机科学与技术]