结构化混合注意力网络的图像超分辨率重建  

Image Super-resolution Reconstruction Based on Structured Fusion Attention Network

作　　者：喻九阳 张德安[1,2] 戴耀南 胡天豪 夏文凤 YU Jiuyang;ZHANG Dean;DAI Yaonan;HU Tianhao;XIA Wenfeng(Hubei Green Chemical Equipment Engineering Technology Research Center,Wuhan 430205,China;School of Mechanical and Electrical Engineering,Wuhan Institute of Technology,Wuhan 430205,China)

机构地区：[1]湖北省绿色化工装备工程技术研究中心,武汉430205 [2]武汉工程大学机电工程学院,武汉430205

出　　处：《计算机科学》2023年第S01期352-356,共5页Computer Science

基　　金：湖北省重点研发计划(2020BAB030)。

摘　　要：针对现有图像超分辨模型存在特征提取能力弱、模型参数量较复杂等问题,提出了一种结构化混合注意力网络的图像超分辨率重建模型,该模型在提高图像超分辨率重建效果的同时降低了模型的参数量。首先,对编码器进行结构化处理,通过通道数量的不同来提取更多的图像特征。其次,对编码器的输出特性进行注意力网络混合重组,从而加强图像的特征特性。最后,采用残差方式将输入的浅层图像特征直接与强化特征相混合,降低网络的参数量。实验结果表明,在公共数据集及不同放大倍率的前提下,文中构建模型的PSNR值和SSIM值基本是最优的,且网络结构的参数量较低,较好地平衡了图像超分辨率重建过程中性能和参数复杂度间的关系。Aiming at the problems of weak feature extraction ability and complex model parameters in existing image super-resolution models,an image super-resolution reconstruction model based on structured hybrid attention network is proposed.This model can reduce the number of parameter while improving the super-resolution reconstruction effect.First,the encoder is structured to extract more image features through the difference in the number of channels.Second,the attention network hybrid reorganization is performed on the output features of the encoder to enhance the feature characteristics of the image.Finally,a residual method is used to directly mix the input shallow image features with the enhanced features to reduce the amount of network parameters.Experimental results show that under the premise of public data sets and different magnifications,the PSNR value and SSIM value of the proposed model are basically optimal,and the parameter amount of the network structure is low,which better balances the relationship between performance and parameter complexity in the process of image super-resolution reconstruction.

关 键 词：图像处理 超分辨率 结构化残差 混合注意力 低模型参数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]