端到端优化的视频编解码技术研究  

Research on end-to-end optimized video encoding and decoding technology

作　　者：吴胤鹏 WU Yinpeng(Nanjing Panda Communications Technology Co.,Ltd.,Nanjing 210000,China)

机构地区：[1]南京熊猫通信科技有限公司,南京210000

出　　处：《计算机应用文摘》2025年第3期186-187,190,共3页

摘　　要：文章深入探讨了端到端优化的视频编解码技术,通过整合视频采集、编码、传输、解码等环节,该技术可以实现对视频数据的高效压缩与高质量传输。随着深度学习技术的发展,基于深度学习的端到端视频编解码方法逐渐兴起,依托神经网络模型强大的学习能力,实现了对传统视频编解码算法的超越。文章首先介绍了端到端视频编解码技术的背景与意义,随后详细阐述了其关键技术,包括基于深度学习的视频压缩算法、多尺度特征的运动补偿、多参考帧的辅助预测帧生成等。通过对比实验,文章展示了端到端优化视频编解码技术在提高压缩效率、减少传输延迟、提升视频质量等方面的显著优势,同时总结了当前研究的成果,并对未来研究方向进行了展望。This paper deeply discusses the end-to-end optimized video codec technology,which can realize efficient compression and high-quality transmission of video data by integrating video capture,encoding,transmission and decoding.With the development of deep learning technology,end-to-end video codec methods based on deep learning have gradually emerged.Relying on the powerful learning ability of neural network models,they have surpassed traditional video codec algorithms.This paper first introduces the background and significance of end-to-end video codec technology,and then elaborates its key technologies,including video compression algorithm based on deep learning,motion compensation of multi-scale features,and auxiliary prediction frame generation of multi-reference frames.Through comparative experiments,this paper demonstrates the significant advantages of end-to-end optimized video codec technology in improving compression efficiency,reducing transmission delay,and improving video quality.At the same time,the current research results are summarized,and the future research direction is prospeced.

关 键 词：端到端优化 视频编解码 深度学习 多尺度特征 多参考帧 压缩效率 传输延迟 视频质量 

分 类 号：TN919[电子电信—通信与信息系统]