基于非对称特征差异网络的图像超分辨率重建  

Asymmetric feature difference network for image super⁃resolution reconstruction

作　　者：王彩玲 沈齐 WANG Cailing;SHEN Qi(College of Automation&College of Artificial Intelligence,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学自动化学院、人工智能学院,江苏南京210023

出　　处：《南京邮电大学学报（自然科学版）》2022年第2期79-84,共6页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：南京邮电大学自然科学基金(NY220057)资助项目。

摘　　要：在众多图像处理任务中,图像的质量好坏对任务的影响是巨大的。对于低分辨率图像带来的像素模糊和细节缺失等问题,提出一种基于非对称特征差异网络模型用于图像超分辨率重建,将自监督的对比学习思想应用在无监督的图像超分辨率重建领域。通过对比两个网络之间特征图的差异,计算出损失并反向更新网络的权重,增强网络学习图像内部数据分布的能力。设计了一种特征增强模块,进一步提高生成结果的质量。经过图像质量评价指标的客观实验和可视化结果的主观展示,提出的方法能够有效地提高细节特征和图像质量。Qualities of images are crucial to image processing tasks.For the problems of blurred pixels and missing details caused by low⁃resolution images,this paper proposes an asymmetric feature difference network for image super⁃resolution,and applies the self⁃supervised contrast learning idea to unsupervised image super⁃resolution.The differences of the feature maps between the two networks are identified,and then the loss is calculated and the weight of the network is updated in the reverse direction.This enhances the network􀆳s ability of learning the internal data distribution of the image.A feature enhancement module is designed to further improve the quality of the generated results.The results of the evaluation experiments and visualization demonstrate that the proposed method can effectively improve the detailed features and image quality.

关 键 词：对比学习 非对称孪生网络 特征提取与特征差异 超分辨率重建 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]