融合小波卷积的轻量化图像超分辨率重建

doi:10.19678/j.issn.1000-3428.0252333

摘要/Abstract

摘要： 单图像超分辨率（SISR）任务当前面临的主要难题在于，如何在恢复低分辨率（LR）图像的高频细节同时降低计算复杂度和避免伪影的产生。为了有效解决这些问题，提出融合小波卷积的多尺度大核网络WEMA-Net。该网络首先把小波变换卷积(WTConv)与大核分解策略结合，在频域内进行卷积操作，扩展感受野并降低计算复杂度。再设计多尺度分组机制，在不同尺度上捕获长距离依赖和局部纹理信息，结合小波卷积减少大核卷积可能导致的伪影。接着提出自适应门控空间注意力单元，根据不同区域的特征动态调整注意力权重，从而优化局部和全局特征的融合，增强细节和边缘的恢复能力。此外，设计调制卷积单元模块，通过提升对异常值的鲁棒性，进一步提高特征提取的灵活性。在Set5、Set4、B100、Urban100、Manga109等数据集上的实验结果表明，WEMA-Net在峰值信噪比（PSNR）上相较于轻量化模型RFDN、BSRN、LKDN、RepRFN、OSFFNet等均有提升。在Urban100和Manga109数据集上，4倍超分辨率性能相比于LKDN分别提高0.18 dB和0.20 dB。实验结果验证了该网络在图像细节恢复、计算效率及鲁棒性方面的优越性，表明该方法具有广泛的应用前景。

关键词: 单图像超分辨率, 深度学习, 小波变换卷积, 注意力机制, 轻量化

Abstract: The primary challenge of Single Image Super-Resolution (SISR) tasks is to recover the high-frequency details of low-resolution images while reducing computational complexity and avoiding artifacts. To solve these problems effectively, a multiscale large-kernel network, WEMA-Net, fused with wavelet convolution, is proposed. First, the network combines wavelet transform convolution with a large kernel decomposition strategy and performs convolution operations in the frequency domain, which expands the receptive field and reduces computational complexity. Subsequently, a multiscale grouping mechanism is introduced to capture long-distance dependence and local texture information at different scales, and wavelet convolution is combined to reduce the artifacts that may be caused by large kernel convolution. Subsequently, an adaptive gated spatial attention unit is proposed that can dynamically adjust the attention weight according to the features of different regions for optimizing the fusion of local and global features and enhancing the recovery ability of details and edges. Additionally, a modulation convolution unit module is introduced, which further improves the flexibility of feature extraction by improving the robustness to outliers. Experiments on datasets such as Set5, Set14, B100, Urban100, and Manga109 reveal that WEMA-Net improves the Peak Signal-to-Noise Ratio (PSNR) compared to lightweight models such as RFDN, BSRN, LKDN, RepRFN, and OSFFNet. In the 4×super-resolution, the proposed method achieves 0.18 dB and 0.20 dB higher PSNR than LKDN on Urban100 and Manga109, respectively. The experimental results verify the superiority of the network in image detail recovery, computational efficiency, and robustness, indicating that the proposed method has wide application prospects.

Key words: Single Image Super-Resolution (SISR), deep learning, Wavelet Transform convolution (WTConv), attention mechanism, lightweight

中图分类号:

TP391

胡静, 黄文韬, 贺秉毅. 融合小波卷积的轻量化图像超分辨率重建[J]. 计算机工程, 2026, 52(11): 239-249.

HU Jing, HUANG Wentao, HE Bingyi. Lightweight Image Super-Resolution Reconstruction Fused with Wavelet Convolution[J]. Computer Engineering, 2026, 52(11): 239-249.

https://www.ecice06.com/CN/Y2026/V52/I11/239

参考文献

[1] 赖厚增. 基于特征增强的深度残差超分辨率重建网络研究[D]. 南昌: 江西财经大学, 2023.LAI H Z. Research on deep residual super-resolution reconstruction network based on feature enhancement[D]. Nanchang: Jiangxi University of Finance and Economics, 2023. (in Chinese)
[2] 胡波, 龚兵兵, 李纯熠, 等. 高频恢复和引导的两阶段图像去模糊[J]. 重庆邮电大学学报(自然科学版), 2025, 37(6): 931-939. HU B, GONG B B, LI C Y, et al. Two-stage image deblurring with high-frequency recovery and guidance[J]. Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition), 2025, 37(6): 931-939. (in Chinese)
[3] 郭琳, 刘坤虎, 马晨阳, 等. 基于感受野扩展残差注意力网络的图像超分辨率重建[J]. 计算机应用, 2024, 44(5): 1579-1587. GUO L, LIU K H, MA C Y, et al. Image super-resolution reconstruction based on residual attention network with receptive field expansion[J]. Journal of Computer Applications, 2024, 44(5): 1579-1587. (in Chinese)
[4] 黄峰, 刘鸿伟, 沈英, 等. 基于多尺度空间自适应注意力网络的轻量级图像超分辨率方法[J]. 模式识别与人工智能, 2025, 38(1): 36-50. HUANG F, LIU H W, SHEN Y, et al. Lightweight image super-resolution reconstruction method based on multi-scale spatial adaptive attention network[J]. Pattern Recognition and Artificial Intelligence, 2025, 38(1): 36-50. (in Chinese)
[5] 张锡英, 王世宇, 李金凤, 等. 轻量级多尺度大核卷积的图像超分辨率重建网络[J]. 北京邮电大学学报, 2025, 48(2): 73-79. ZHANG X Y, WANG S Y, LI J F, et al. Lightweight multi-order gated aggregation network for image super-resolution[J]. Journal of Beijing University of Posts and Telecommunications, 2025, 48(2): 73-79. (in Chinese)
[6] AHN N, KANG B, SOHN K A. Fast, accurate, and lightweight super-resolution with cascading residual network[C]//Proceedings of the 15th European Conference on Computer Vision. Berlin, Germany: Springer, 2018: 256-272.
[7] 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, USA: ACM Press, 2019: 2024-2032.
[8] LIU J, TANG J, WU G S. Residual feature distillation network for lightweight image super-resolution[C]//Proceedings of the 16th European Conference on Computer Vision. Berlin, Germany: Springer, 2020: 41-55.
[9] LI Z Y, LIU Y Q, CHEN X Y, et al. Blueprint separable residual network for efficient image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, USA: IEEE Press, 2022: 832-842.
[10] MAO Y Y, ZHANG N H, WANG Q, et al. Multi-level dispersion residual network for efficient image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada: IEEE Press, 2023: 1660-1669.
[11] XIE C X, ZHANG X M, LI L Z, et al. Large kernel distillation network for efficient single image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada: IEEE Press, 2023: 1283-1292.
[12] 张豪, 马冀, 袁江. 基于特征层次递进融合的轻量级图像超分辨率网络[J]. 计算机系统应用, 2025, 34(1): 118-127. ZHANG H, MA J, YUAN J. Lightweight image super-resolution network based on hierarchical progressive fusion of feature[J]. Computer Systems & Applications, 2025, 34(1): 118-127. (in Chinese)
[13] FINDER S E, AMOYAL R, TREISTER E, et al. Wavelet convolutions for large receptive fields[C]//Proceedings of European Conference on Computer Vision. Berlin, Germany: Springer, 2024: 363-380.
[14] WANG Y, LI Y S, WANG G, et al. Multi-scale attention network for single image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, USA: IEEE Press, 2024: 5950-5960.
[15] ZHENG M J, SUN L, DONG J X, et al. SMFANet: a lightweight self-modulation feature aggregation network for efficient image super-resolution[C]//Proceedings of European Conference on Computer Vision. Berlin, Germany: Springer, 2024: 359-375.
[16] DONG C, LOY C C, HE K M, et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(2): 295-307.
[17] KIM J, LEE J K, LEE K M. Accurate image super-resolution using very deep convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE Press, 2016: 1646-1654.
[18] 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. Honolulu, USA: IEEE Press, 2017: 1132-1140.
[19] 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. Berlin, Germany: Springer, 2018: 286-301.
[20] ASHISH V. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing System. New York, USA: ACM Press, 2017: 6000-6010.
[21] CHEN H T, WANG Y H, GUO T Y, et al. Pre-trained image processing transformer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, USA: IEEE Press, 2021: 12299-12310.
[22] LIANG J Y, CAO J Z, SUN G L, et al. SwinIR: image restoration using swin transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal, Canada: IEEE Press, 2021: 1833-1844.
[23] LI Y W, ZHANG Y L, TIMOFTE R, et al. NTIRE 2023 challenge on efficient super-resolution: methods and results[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2023: 1922-1960.
[24] DONG C, LOY C C, TANG X O. Accelerating the super-resolution convolutional neural network[C]//Proceedings of the 14th European Conference on Computer Vision. Berlin, Germany: Springer, 2016: 391-407.
[25] 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 Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE Press, 2016: 1874-1883.
[26] 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, USA: ACM Press, 2019: 2024-2032.
[27] KONG F Y, LI M X, LIU S W, et al. Residual local feature network for efficient super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, USA: IEEE Press, 2022: 765-775.
[28] LI Y W, ZHANG K, TIMOFTE R, et al. NTIRE 2022 challenge on efficient super-resolution: methods and results[C]//Proceedings of Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2022: 1061-1101.
[29] YU L, LI X P, LI Y W, et al. DIPNet: efficiency distillation and iterative pruning for image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada: IEEE Press, 2023: 1692-1701.
[30] DENG W J, YUAN H J, DENG L H, et al. Reparameterized residual feature network for lightweight image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada: IEEE Press, 2023: 1712-1721.
[31] 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. Menlo Park, USA：AAAI Press, 2023: 2375-2383.
[32] WANG K X, YANG X M, JEON G. Hybrid attention feature refinement network for lightweight image super-resolution in metaverse immersive display[J]. IEEE Transactions on Consumer Electronics, 2024, 70(1): 3232-3244.
[33] WANG Y, ZHANG T. OSFFNet: Omni-stage feature fusion network for lightweight image super-resolution[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park, USA：AAAI Press, 2024, 38(6): 5660-5668.
[34] BEVILACQUA M, ROUMY A, GUILLEMOT C, et al. Low-complexity single-image super-resolution based on nonnegative neighbor embedding[EB/OL].[2024-03-05]. http://eprints.imtlucca.it/2412/1/Bevilacqua_2012.pdf.
[35] ZEYDE R, ELAD M, PROTTER M. On single image scale-up using sparse-representations[C]// Proceedings of the 7th International Conference on Curves and Surfaces. Berlin, Germany: Springer, 2010: 711-730.
[36] MARTIN D, FOWLKES C, TAL D, et al. A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics[C]//Proceedings of the 8th IEEE International Conference on Computer Vision. Vancouver, Canada: IEEE Press, 2002: 416-423.
[37] HUANG J B, SINGH A, AHUJA N. Single image super-resolution from transformed self-exemplars[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston, USA: IEEE Press, 2015: 5197-5206.
[38] MATSUI Y, ITO K, ARAMAKI Y, et al. Sketch-based manga retrieval using Manga109 dataset[J]. Multimedia Tools and Applications, 2017, 76(20): 21811-21838.
[39] ZHANG R, ISOLA P, EFROS A A, et al. The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE Press, 2018: 586-595.
[40] WANG L G, DONG X Y, WANG Y Q, et al. Exploring sparsity in image super-resolution for efficient inference[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, USA: IEEE Press, 2021: 4915-4924.
[41] LI X, DONG J X, TANG J H, et al. DLGSANet: lightweight dynamic local and global self-attention network for image super-resolution[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Paris, France: IEEE Press, 2024: 12746-12755.

选择文件类型/文献管理软件名称

选择包含的内容