基于深度学习的HEVC编码复杂度优化方法  

A deep learning approach for complexity reductionof HEVC

作　　者：沈玉志[1] 金雪莹 李天一[2] SHEN Yuzhi;JIN Xueying;LI Tianyi(School of Business Administration,Liaoning Technical University,Fuxin 123000,China;Electronic Information Engineering,Beihang University,Beijing 100191,China)

机构地区：[1]辽宁工程技术大学工商管理学院,辽宁阜新123000 [2]北京航空航天大学电子信息工程学院,北京100191

出　　处：《辽宁工程技术大学学报（自然科学版）》2024年第3期351-358,共8页Journal of Liaoning Technical University (Natural Science)

摘　　要：为降低高效率视频编码(HEVC)的计算复杂度,提出一种基于深度学习的复杂度优化方法。构建大规模编码单元(CU)划分数据库,为算法设计和神经网络训练提供数据基础;设计一种适用于HEVC的CU划分图模型,高效表征多个邻近CU的划分模式;在此基础上,提出一种基于稠密网络的分层卷积神经网络结构,准确预测HEVC中的三级CU划分结果。实验结果表明:该方法能够在保证编码效率前提下,平均节省57%的编码时间,有效解决了HEVC编码复杂度过高的瓶颈。For reducing the computational complexity of the high efficiency video coding(HEVC)standard,this paper proposes a complexity reduction approach based on deep learning.Firstly,a large-scale database for the coding unit(CU)partition is established,providing fundamental data for algorithm design and neural network training.Then,a CU partition map model is designed adaptive to HEVC,as an efficient representation of the CU partition across multiple adjacent CUs.Based on the model,a DenseNet-based hierarchical convolutional neural network structure is proposed to accurately predict the three-level CU partition in HEVC.The experimental results show that the proposed approach can save 57%of the encoding time on average while maintaining the coding efficiency,considerably alleviating the bottleneck of over-high encoding complexity in HEVC.

关 键 词：高效率视频编码 复杂度优化 编码单元 深度学习 稠密网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]