联合边路和中路解码特征学习的多描述编码图像增强方法  

Multiple description coded image enhancement method with joint learning of side-decoding and central-decoding features

作　　者：赵利军 曹聪颖 张晋京[2] 白慧慧[3] 赵耀[3] 王安红[1] Zhao Lijun;Cao Congying;Zhang Jinjing;Bai Huihui;Zhao Yao;Wang Anhong(College of Electronic Information Engineering,Taiyuan University of Science&Technology,Taiyuan 030024,China;College of Big Science&Technology,North University of China,Taiyuan 030051,China;Institute of Information Science,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]太原科技大学电子信息工程学院,太原030024 [2]中北大学大数据学院,太原030051 [3]北京交通大学信息科学研究所,北京100044

出　　处：《计算机应用研究》2022年第9期2873-2880,共8页Application Research of Computers

基　　金：太原科技大学博士科研启动基金资助项目(20192023);山西省基础研究计划资助项目(202103021223284);来晋工作优秀博士奖励资金资助项目(20192055);太原科技大学研究生教育创新项目(XCX212029);国家自然基金资助项目(61972023,62072325)。

摘　　要：提出一种联合边路和中路解码特征学习的多描述编码图像增强方法。该方法同时考虑了边路解码图像增强和中路解码图像增强的问题,因而可以通过联合学习优化中路解码和边路解码的特征来实现更好的网络训练。首先,考虑到多描述编码的边路独立解码和中路联合解码的特性,提出一种网络共享的边路低分辨率特征提取网络来有效地提取具有相同内容和差异细节的两个边路解码图像的特征,同时设计一种残差递归补偿网络结构并将其用于边路与中路低分辨率特征提取网络。其次,设计一种多描述边路上采样重建网络,该网络采用部分网络层参数共享策略,该策略能够减小网络模型参数量,同时提高网络的泛化能力。最后,提出一种多描述中路上采样重建网络,将两个边路低分辨率特征与中路低分辨率特征进行深层特征融合来实现多描述压缩图像的增强。大量的实验结果表明:在模型复杂度、客观质量和视觉质量评价方面,所提方法优于很多的图像增强方法如ARCNN、FastARCNN、DnCNN、WSR和DWCNN。This paper proposed MDC image enhancement method by using joint learning of side-decoding and central-decoding features,which considered the problems of side decoding image enhancement and central decoding image enhancement at the same time,so it could realize better network training by optimizing central decoding and side decoding features through joint learning.First,considering side independent decoding and central joint decoding features for MDC,this paper proposed a network-sharing side low-resolution feature extraction network to effectively extract features from two-side decoded images with the same content and different details,while designed a residual recursive compensation network structure and applied it into both side and central low-resolution feature extraction network.Secondly,it designed a multiple description up-sampling reconstruction network,which adopted parameter sharing strategy for partial layers of network,which could reduce parameter number of network model and improve network generalization ability.Finally,it proposed multiple description central up-sampling reconstruction network to perform deep feature fusion with two low-resolution side features and central features to enhance multiple description compressed images.A large number of experimental results show that the proposed method is superior to several image enhancement methods such as ARCNN,FastARCNN,DnCNN,WSR and DWCNN in terms of model complexity,objective quality and visual quality assessment.

关 键 词：多描述编码 深度学习 图像增强 压缩失真 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]