融合边缘增强与非局部模块的遥感图像超分辨率重建生成对抗网络  

Generate adversarial network for super-resolution reconstruction of remote sensing images by fusing edge enhancement and non-local modules

作　　者：刘杰 祁箬 韩轲[2] LIU Jie;QI Ruo;HAN Ke(College of Measurement and Control Technology and Communication Engineering,Harbin University of Science and Technology,Harbin 150080,China;School of Computer and Information Engineering,Harbin University of Commerce,Harbin 150028,China)

机构地区：[1]哈尔滨理工大学测控技术与通信工程学院,黑龙江哈尔滨150080 [2]哈尔滨商业大学计算机与信息工程学院,黑龙江哈尔滨150028

出　　处：《光学精密工程》2023年第14期2080-2092,共13页Optics and Precision Engineering

基　　金：黑龙江省自然科学基金资助项目(No.LH2020F009)。

摘　　要：针对遥感图像成像过程中噪声污染严重,超分辨率重建图像存在目标边缘模糊和伪影等问题,本文提出一种融合边缘增强与非局部模块的遥感图像超分辨率算法(Edge-Enhanced and Non-local Modules Generative Adversarial Net⁃work,ENGAN)。为了使图像细节边缘更清晰,本文融合一种图像边缘增强模块;同时,为进一步扩大模型感受野和增强去除边缘噪声性能,改进边缘增强模块中的Mask分支;此外,引入非局部模块,通过更好地利用图像的内在特征相关性,进一步提升了网络的重建性能。本文在UCAS-AOD和NWPU VHR-10两种遥感图像数据集上进行多个算法的对比实验,结果表明本文提出的方法在多个评价指标上均有所改善。以退化类型Ⅳ为例,本文方法相比深度盲超分辨率退化模型,4倍超分辨率的SSIM提升了0.068,PSNR提升了1.400 dB,RMSE减少了12.5%,且重建后的遥感图像相较于原始图像可以得到更好的地面目标检测结果。To address the serious noise pollution in the process of image remote sensing and the existence of object edge blur and artifacts in the super-resolution reconstructed image,this study proposes a remote sensing image super-resolution algorithm called edge-enhanced and non-local modules generative adversari⁃al network(ENGAN).To make the image edge details clearer,the proposed algorithm integrated an im⁃age edge enhancement module.To further expand the receptive field of the model and enhance the edge noise removal,the Mask branch in the edge enhancement module was simultaneously improved.The use of the intrinsic feature correlation of images further improved the reconstruction performance of the net⁃work.In this study,comparison experiments of multiple algorithms were performed on two remote sens⁃ing image datasets,UCAS-AOD and NWPU VHR-10.The proposed method showed improvement in multiple evaluation indicators.Taking degradation type IV as an example,the 4x super-resolution SSIM was increased by 0.068,PSNR increased by 1.400 dB,and RMSE reduced by 12.5%compared with the deep-blind super-resolution degradation model.Moreover,the reconstructed remote sensing image can ob⁃tain better ground target detection results than the original image.

关 键 词：遥感图像 超分辨率 边缘增强 非局部特征 生成对抗网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]