Multi-perception large kernel convnet for efficient image super-resolution  

基于多感知大核卷积的轻量级图像超分辨率重建方法

作　　者：MIAO Xuan LI Zheng XU Wen-Zheng 缪炫;李征;徐文政(四川大学计算机学院(软件学院),成都610065;四川大学天府工程数值模拟与软件创新中心,成都610207)

机构地区：[1]College of Computer Science(College of Software),Sichuan University,Chengdu 610065,China [2]Tianfu Engineering-Oriented Numerical Simulation&Software Innovation Center,Sichuan University,Chengdu 610207,China

出　　处：《四川大学学报(自然科学版)》2025年第1期67-78,共12页Journal of Sichuan University(Natural Science Edition)

基　　金：国家重点研发计划项目(2020YFA0714003);国家重大项目(GJXM92579);四川省科技厅重点研发项目(2021YFQ0059)。

摘　　要：Significant advancements have been achieved in the field of Single Image Super-Resolution(SISR)through the utilization of Convolutional Neural Networks(CNNs)to attain state-of-the-art performance.Recent efforts have explored the incorporation of Transformers to augment network performance in SISR.However,the high computational cost of Transformers makes them less suitable for deployment on lightweight devices.Moreover,the majority of enhancements for CNNs rely predominantly on small spatial convolutions,thereby neglecting the potential advantages of large kernel convolution.In this paper,the authors propose a Multi-Perception Large Kernel convNet(MPLKN)which delves into the exploration of large kernel convolution.Specifically,the authors have architected a Multi-Perception Large Kernel(MPLK)module aimed at extracting multi-scale features and employ a stepwise feature fusion strategy to seamlessly integrate these features.In addition,to enhance the network's capacity for nonlinear spatial information processing,the authors have designed a Spatial-Channel Gated Feed-forward Network(SCGFN)that is capable of adapting to feature interactions across both spatial and channel dimensions.Experimental results demonstrate that MPLKN outperforms other lightweight image super-resolution models while maintaining a minimal number of parameters and FLOPs.在单图像超分辨率的研究领域,卷积神经网络已经取得了显著进展,并在性能上达到了行业领先水平.近期的学术研究加大了对Transformers在单图像超分辨率应用的探索力度,以期进一步提升网络性能.尽管如此,由于Transformers的高计算成本,其在轻量级设备上的应用受到了限制.同时,针对卷积神经网络的轻量级优化多数集中在小空间卷积上,而忽略了大核卷积可能带来的优势.本研究通过深入研究大核卷积,提出了一种新型的多感知大核卷积网络(MPLKN).该网络设计了一个创新的多感知大核(MPLK)模块,旨在有效提取多尺度特征,并通过分步特征融合策略实现这些特征的平滑整合.为了进一步增强网络处理非线性空间信息的能力,还设计了一个空间通道门控前馈网络(SCGFN),该网络能够灵活地适应空间和通道维度上的特征交互.实验结果表明,MPLKN在保持较少参数和FLOPs消耗的同时,相较于其他轻量的图像超分辨模型展现出了更好的性能表现.

关 键 词：Single Image Super-Resolution Lightweight model Deep learning Large kernel 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]