基于频率驱动的多尺度图像超分辨率方法

doi:10.11896/jsjkx.250700046

摘要/Abstract

摘要： 单图像超分辨率是一种旨在从单张低分辨率图像重建高分辨率图像的关键技术,被广泛应用于医学影像增强、卫星遥感、安防监控及数字媒体内容优化等领域。其核心在于恢复图像丢失的高频细节,以提升视觉质量。然而,该任务因其病态逆问题的本质而面临诸多挑战:传统插值方法难以重建复杂细节;基于卷积神经网络的方法虽能提取局部特征,但在全局建模上不足;而Transformer虽擅长长程依赖捕捉,却对高频信息细化能力有限,导致重建图像边缘模糊,纹理失真。为此,提出了一种创新的模型,通过引入小波变换实现多尺度图像分解,优化高频信息提取,并设计三大核心模块:小波细化模块,增强高频细节处理;移位矩形特征增强模块,捕获全局上下文;多尺度小波融合模块,整合高频先验与全局特征。该方法显著提升了纹理和边缘清晰度,兼顾了局部细节与整体一致性。实验结果表明,该模型在 Set5,Set14和BSD100 等基准数据集上超越了现有技术,平均峰值信噪比提升约 0.3 dB,主观视觉质量亦更优。该研究不仅有效解决了高频信息恢复难题,还为单图像超分辨率领域提供了新思路,具有重要的学术与应用价值。

关键词: 单图像超分辨率, Transformer, 卷积神经网络, 小波变换, 高频信息优化, 多尺度处理

Abstract: Single-image super-resolution is a critical technique aimed at reconstructing high-resolution images from single low-resolution inputs.It plays a pivotal role in various fields such as medical image enhancement,satellite remote sensing,security surveillance,and digital media content optimization.The core of SISR lies in restoring the lost high-frequency details to improve visual quality.However,this task faces numerous challenges due to its ill-posed inverse problem nature:traditional interpolation methods struggle to reconstruct complex details;convolutional neural network-based approaches can extract local features but fall short in global modeling;while Transformers excel at capturing long-range dependencies,they have limited capabilities for refining high-frequency information,leading to blurred edges and distorted textures in reconstructed images.To address these issues,this study proposes an innovative model that leverages wavelet transformation for multi-scale image decomposition,optimizing the extraction of high-frequency information.Three key modules are designed:the wavelet refinement module enhances the processing of high-frequency details;the shifted rectangular feature enhancement module captures global context;and the multi-scale wavelet fusion module integrates high-frequency priors with global features.This method significantly improves texture clarity and edge sharpness,balancing both local details and overall consistency.Experimental results demonstrate that the proposed model outperforms existing techniques on benchmark datasets such as Set5,Set14,and BSD100,achieving an average peak signal-to-noise ratio improvement of approximately 0.3 dB,along with superior subjective visual quality.This research not only effectively tackles the challenge of high-frequency information recovery but also provides new insights into the field of single-image super-resolution,holding significant academic and practical value.

Key words: Single-image super-resolution, Transformer, Convolutional neural networks, Wavelet transform, High-frequency information optimization, Multi-scale processing

中图分类号:

TP183

杨红菊, 张子扬, 李尧. 基于频率驱动的多尺度图像超分辨率方法[J]. 计算机科学, 2026, 53(5): 218-227. https://doi.org/10.11896/jsjkx.250700046

YANG Hongju, ZHANG Ziyang, LI Yao. Frequency Driven Multi-scale Image Super-resolution Method[J]. Computer Science, 2026, 53(5): 218-227. https://doi.org/10.11896/jsjkx.250700046

参考文献

[1]MUDUNURI S P,BISWAS S.Low resolution face recognition across variations in pose and illumination[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,38(5):1034-1040.
[2]WAN X F,HAN F,JIANG X Q,et al.Research on super-resolution restoration of video surveillance images[J].Laser Journal,2014,35(3):5-8.
[3]GREENSPAN H.Super-resolution in medical imaging[J].The Computer Journal,2009,52(1):43-63.
[4]MENG F,WU S,LI Y,et al.Single remote sensing image super-resolution via a generative adversarial network with stratified dense sampling and chain training[J].IEEE Transactions on Geoscience and Remote Sensing,2023,62:1-22.
[5]JIANG N,WANG L.Quantum image scaling using nearestneighbor interpolation[J].Quantum Information Processing,2015,14:1559- 1571.
[6]KIRKLAND E J.Advanced Computing in Electron Microscopy[M].Springer,2010:261-263.
[7]GAO S K,GRUEV V.Bilinear and bicubic interpolation me-thods for division of focal plane polarimeters[J].Optics Express,2011,19(27):26161- 26173.
[8]KIM JW,LEE J K,LEE K M.Accurate image super-resolutionusing very deep convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2016:1646- 1654.
[9]LEDIG C,THEIS L,HUSZÁR F,et al.Photo-realistic single image super-resolution using a generative adversarial network[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2017:4681- 4690.
[10]HAN K,WANG Y H,CHEN H T,et al.A survey on vision transformer[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2022,45(1):87-110.
[11]ZHANG D S.Wavelet transform[M].Cham:Springer,2019:35-44.
[12]SUNKAVALLI K,JOHNSON M K,MATUSIK W,et al.Multi-scale image harmonization[J].ACM Transactions on Graphics,2010,29(4):1-10.
[13]DONG C,LOY C C,HE K M,et al.Learning a deep convolutional network for image super-resolution[C]//Computer Vision-ECCV 2014:13th European Conference.Springer,2014:184-199.
[14]ZHANG Y L,LI K P,LI K,et al.Image super-resolution using very deep residual channel attention networks[C]//Proceedings of the European Conference on Computer Vision(ECCV).Cham:Springer,2018:286-301.
[15]HUI Z,GAO X B,YANG Y C,et al.Lightweight image super-resolution with information multi-distillation network[C]//Proceedings of the 27th ACM International Conference on Multimedia.Cham:Springer,2019:2024-2032.
[16]LIU J,TANG J,WU G S.Residualfeature distillation network for lightweight image super-resolution[C]//Proceedings of Computer Vision-ECCV 2020 Workshops.Cham:Springer,2020:41-55.
[17]ZHOU S C,ZHANG J W,ZUO W M,et al.Cross-scale internal graph neural network for image super-resolution[J].Advances in Neural Information Processing Systems,2020,33:3499-3509.
[18]NIU B,WEN W L,REN W Q,et al.Single image super-resolution via a holistic attention network[C]//Proceedings of Computer Vision-ECCV 2020:16th European Conference.Cham:Springer,2020:191-207.
[19]VASWANI A,SHAZEER N,PARMAR N,et al.Attention isall you need[C]//Advances in Neural Information Processing Systems.2017.
[20]LIANG J Y,CAO J Z,SUN G L,et al.SwinIR:Image restoration usingswin transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Piscataway,NJ:IEEE,2021:1833-1844.
[21]DONG X Y,BAO J M,CHEN D D,et al.Cswin transformer:a general vision transformer backbone with cross-shaped windows[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2022:12124-12134.
[22]CHEN X Y,WANG X T,ZHOU J T,et al.Activating more pixels in image super-resolution transformer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2023:22367-22377.
[23]WOO S,PARK J,LEE J Y,et al.Cbam:Convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision(ECCV).Cham:Springer,2018:3-19.
[24]LI A,ZHANG L,LIU Y,et al.Feature modulation transformer:Cross-refinement of global representation via high-frequency prior for image super-resolution[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Piscataway,NJ:IEEE,2023:12514-12524.
[25]MALLAT S G.Multifrequency channel decompositions of images and wavelet models[J].IEEE Transactions on Acoustics,Speech,and Signal Processing,2002,37(12):2091-2110.
[26]DUAN L J,WU C L,EN Q,et al.Image Super-Resolution Algorithm of Deep Residual Network Based on Wavelet Domain[J].Journal of Software,2019,30(4):941-953.
[27]HUANG H B,HE R,SUN Z N,et al.Wavelet-srnet:Awavelet-based cnn for multi-scale face super resolution[C]//Procee-dings of the IEEE International Conference on Computer Vision.Piscataway,NJ:IEEE,2017:1689-1697.
[28]MA C Y,ZHU J W,LI Y J,et al.Single image super resolution via wavelet transform fusion and SRFeat network[J].Journal of Ambient Intelligence and Humanized Computing,2022,13(11):1-9.
[29]DHAREJO F A,DEEBA F,ZHOU Y C,et al.Twist-gan:towards wavelet transform and transferred gan for spatio-temporal single image super resolution[J].ACM Transactions on Intelligent Systems and Technology,2021,12(6):1-20.
[30]SHI W Z,CABALLERO J,HUSZÁR F,et al.Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2016:1874-1883.
[31]WANG W,SUN Y J,WANG X.Lightweight frequency and spatial feature fused multi-scale remote sensing sceneclassification network[J].Journal of Jilin University(Engineering and Technology Edition),2025,55(10):3361-3371.
[32]骆迪,张乾.基于多尺度注意力的图像超分辨研究[J/OL].重庆工商大学学报(自然科学版),1-14[2025-12-08].https://link.cnki.net/urlid/50.1155.N.20240926.0921.002.
[33]ZHONG M Y,JIANG L,LI C.Image super-resolution reconstruction algorithm based on multi-scale feature extraction residual network[J].Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition),2024,36(1):76-85.
[34]FINDER S E,AMOYAL R,TREISTER E,et al.Wavelet convolutions for large receptive fields[C]//Proceedings of the European Conference on Computer Vision.Cham:Springer,2024:363-380.
[35]WANG W X,CHEN W,QIU Q B,et al.Crossformer++:A versatile vision transformer hinging on cross-scale attention[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,46(5):3123-3136.
[36]ZAMIR S W,ARORA A,KHAN S,et al.Restormer:Efficient transformer for high-resolution image restoration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2022:5728-5739.
[37]LIM B,SON S,KIM H,et al.Enhanced deep residual networks for single image super-resolution[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.Piscataway,NJ:IEEE,2017:136-144.
[38]LI W B,ZHOU K,QI L,et al.Lapar:Linearly-assembled pixel-adaptive regression network for single image super-resolution and beyond[J].Advances in Neural Information Processing Systems,2020,33:20343-20355.
[39]LUO X T,XIE Y,ZHANG Y L,et al.Latticenet:Towardslightweight image super-resolution with lattice block[C]//Proceedings of Computer Vision-ECCV 2020:16th European Conference.Cham:Springer,2020:272-289.
[40]HUI Z,GAO X B,YANG Y C,et al.Lightweight image super-resolution with information multi-distillation network[C]//Proceedings of the 27th ACM International Conference on Multimedia.New York:ACM,2019:2024-2032.
[41]AHN N,KANG B,SOHN K A.Fast,accurate,and lightweight super-resolution with cascading residual network[C]//Procee-dings of the European Conference on Computer Vision(ECCV).Cham:Springer,2018:252-268.
[42]SUN B,ZHANG Y L,JIANG S Y,et al.Hybrid pixel-unshuffled network for lightweight image super-resolution[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Washington,DC:AAAI,2023:2375-2383.
[43]LU Z S,LI J C,LIU H,et al.Transformer for single image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2022:457-466.
[44]RAY A,KUMAR G,KOLEKAR M H.CFAT:Unleashing Triangular Windows for Image Super-Resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2024:26120-26129.
[45]NI Z K,ZHANG Y,YANG W H,et al,Structural Similarity-Inspired Unfolding for Lightweight Image Super-Resolution[J].IEEE Transactions on Image Processing,2025,34:3861-3872.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed