多级信息补偿的U型网络图像超分辨率重建算法  

Image super-resolution reconstruction algorithm of U-shaped network based on multi-level information compensation

作　　者：宋昭漾 赵小强[1,2,3] 惠永永 蒋红梅[1,2,3] SONG Zhao-yang;ZHAO Xiao-qiang;HUI Yong-yong;JIANG Hong-mei(College of Electrical Engineering and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou 730050,China;National Experimental Teaching Center of Electrical and Control Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学电气工程与信息工程学院,兰州730050 [2]甘肃省工业过程先进控制重点实验室,兰州730050 [3]兰州理工大学国家级电气与控制工程实验教学中心,兰州730050

出　　处：《控制与决策》2023年第9期2479-2486,共8页Control and Decision

基　　金：国家自然科学基金项目(61763029);国防基础科研项目(JCKY2018427C002);国家重点研发计划项目(2020YFB1713600);甘肃省重点研发计划项目(21YF5GA072);甘肃省自然科学基金项目(21JR7RA206,20JR5RA459);甘肃省教育厅优秀研究生创新之星项目(2021CXZX-501)。

摘　　要：针对基于深度神经网络的图像超分辨率重建算法在特征提取过程中容易丢失特征信息,导致重建图像缺少纹理和边缘细节等问题,提出一种多级信息补偿的U型网络图像超分辨率重建算法.首先设计一个用于图像超分辨率重建的U型网络,该网络通过下通道分支对输入特征进行多层级特征提取和通道压缩,通过底层模块对压缩后的特征进行融合并提取不同通道的相关特征,通过上通道分支对压缩后的相关特征进行多层次特征提取和通道恢复;然后设计多级信息补偿模型,对U型网络的通道压缩过程中丢失的信息和通道恢复过程中难以恢复的信息进行补偿;最后在不同放大倍数下的Set5、Set14、BSD100和Urban100测试集上对所提算法和主流算法进行对比测试分析,实验结果表明所提算法相比主流算法实现了在峰值信噪比(PSNR)/结构相似度(SSIM)指标和视觉效果上的巨大提升.Aiming at the problem that the image super-resolution reconstruction algorithm based on the deep neural network is easy to lose feature information during feature extraction,which leads to the lacks of texture and edge details in the reconstructed image,this paper proposes a U-shaped network image super-resolution reconstruction algorithm with multi-level information compensation.Firstly,a U-shaped network for image super-resolution reconstruction is designed,which performs multi-level feature extraction and channel compression on the input features through the lower channel branch,and fuses the compressed features through the bottom module and extracts the related features of different channels,and performs multi-level feature extraction and channel recovery on the compressed related features through the upper channel branch.Then,a multi-level information compensation model is designed,which compensates for the lost information in the channel compression process and the information that is difficult to recover in the channel recovery process of the U-shaped network.Finally,compared with the mainstream algorithms,the proposed algorithm is tested and analyzed by the test sets of Set5,Set14,BSD100 and Urban100 at different magnifications.The experimental results show that the proposed algorithm greatly improves the peak signal-to-noise ratio(PSNR)/structural similarity(SSIM)index and visual effect.

关 键 词：超分辨率重建 深度神经网络 多级信息补偿 U型网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]