基于深度学习的视频压缩方法优化策略  

Optimization Strategy of Video Compression Methods Based on Deep Learning

作　　者：全萌凯 杨振玉 QUAN Mengkai;YANG Zhenyu(Henan College of Surveying and Mapping,Zhengzhou 450000,China)

机构地区：[1]河南测绘职业学院,河南郑州450000

出　　处：《电视技术》2024年第8期27-29,34,共4页Video Engineering

摘　　要：随着视频数据在互联网和多媒体中的广泛应用,高效的视频压缩技术变得尤为重要。为此,研究基于深度学习的视频压缩方法,特别是自编码器结构的应用及优化。首先,详细介绍基于自编码器的视频压缩和解压缩的基本原理,阐述编码器和解码器在处理过程中的功能和作用。其次,针对传统均方误差损失函数在视频质量评估中的不足,引入感知损失函数,结合高级特征映射来更好地评估视频质量,从而优化压缩算法。为了验证优化策略的有效性,采用腾讯视频数据集进行实验,结果表明提出的方法在视频质量保持方面具有显著优势。With the wide application of video data in internet and multimedia applications,efficient video compression technology becomes particularly important.This article investigates video compression methods based on deep learning,particularly the application and optimization of autoencoder structures.Firstly,this article provides a detailed introduction to the basic principles of video compression and decompression based on autoencoders,emphasizing the functions and roles of encoders and decoders in the processing process.Secondly,in response to the shortcomings of the traditional mean square error loss function in video quality evaluation,this study introduces the perceptual loss function,which combines advanced feature mapping to better evaluate video quality and optimize compression algorithms.To verify the effectiveness of the optimization strategy,this paper conducted experiments on the Tencent video dataset,and the results showed that the proposed method has significant advantages in maintaining video quality.

关 键 词：深度学习 自编码器 视频压缩 解压 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]