一种基于编码单元快速划分的VVC帧内编码方法  

A VVC intra coding method based on fast partition for coding unit

作　　者：钟辉 陆宇[1] 殷海兵 黄晓峰 ZHONG Hui;LU Yu;YIN Haibing;HUANG Xiaofeng(College of Communication Engineering,Hangzhou Danzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018

出　　处：《电信科学》2024年第8期23-33,共11页Telecommunications Science

基　　金：国家自然科学基金资助项目(No.61972123);科技部重点研发课题项目(No.2023YFB4502800);浙江省教育厅科研项目(No.Y202249588)。

摘　　要：相比于高效视频编码(high efficiency video coding,HEVC)标准,新一代编码标准多功能视频编码(versatile video coding,VVC)引入了很多新的技术,其中包括四叉树(quadtree,QT)和多类型树(multi-type tree,MTT)划分,MTT划分由HEVC中的QT划分延伸而来。新划分方法提高了压缩效率,但导致编码时间急剧增加。为了降低编码复杂度,提出了一种结合深度学习方法和MTT方向早期判决的快速帧内编码算法。首先使用轻量级的卷积神经网络(convolutional neural network,CNN)对QT和部分MTT进行预测划分,其余MTT则采用提前预测MTT划分方向的方法作进一步的优化。实验结果表明,所提方法能够大幅降低编码复杂度,相比于原始编码器的编码时间减少了74.3%,且只有3.3%的码率损失,性能优于对比的方法。Compared to the high efficiency video coding(HEVC)standard,the latest generation coding standard,versatile video coding(VVC)has introduced many new technologies,including quadtree(QT)and multi-type tree(MTT)partitioning.MTT partition is extended from QT partition in HEVC.The new partition method increases encoding complexity,leading to a sharp increase in encoding time.To reduce encoding complexity,a fast intra coding method combining deep learning methods and early decision in the MTT direction was proposed.Firstly,a lightweight convolutional neural network(CNN)network was used to predict partition for QT and partial MTT.Then,an early prediction for MTT partition direction method was adopted for further optimization of residual MTT.Experimental results show that the proposed method can significantly reduce encoding complexity,with a 74.3%reduction in encoding time compared to the original encoder with only 3.3%rate loss.Moreover,the performance of proposed method is superior to other comparative algorithms.

关 键 词：VVC 帧内编码 卷积神经网络 快速编码 四叉树 多类型树 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]