Superpixel Guide for Transformer Low-Light Image Denoising Method

doi:10.19678/j.issn.1000-3428.0070426

Abstract

Abstract: Existing low-light image denoising methods mainly use the feature extraction and denoising mechanisms of Transformer and Convolutional Neural Networks (CNN). They face two problems: the self attention mechanism based on local windows fails to fully capture the nonlocal self-similarity in images, and the calculation of self-attention in the channel dimension does not fully utilize the spatial correlation of images. To address these issues, this study proposes a superpixel guided strategy for a window partition-based visual Transformer method; the strategy can adaptively select relevant windows for global interactions. First, a Top-N Cross Attention mechanism (TNCA) is designed based on window interactions, the top N windows that are most similar to the target image window are selected dynamically, and the information related to the image windows in the channel dimension are aggregated, fully considering the nonlocal self-similarity of the image. Second, through superpixel segmentation guidance, the expressive power of local features within the window is significantly improved while enhancing the correlation of spatial features in the channel dimension. Finally, a hierarchical Adaptive Interaction Superpixel Guide Transformer (AISGFormer) is constructed. Experimental results show that AISGFormer achieves a Peak Signal-to-Noise Ratio (PSNR) of 39.98 dB and 40.06 dB on the SIDD and DND real image datasets, respectively. Compared with other advanced networks, the PSNR improves by 0.02 dB—14.33 dB and 0.02 dB—7.63 dB, respectively. AISGFormer interacts with local and global information and details more effectively, and it adaptively utilizes self-similarity to suppress region similarity noise.

Key words: low-light image denoising, Transformer, cross-attention, non-local self-similarity, real image noise, superpixel

摘要： 现有的低光图像去噪方法主要使用Transformer和卷积神经网络(CNN)的特征提取和去噪机制,会面临两个问题:基于局部窗口的自注意力机制未能充分捕捉图像中的非局部自相似性;通道维度上的自注意力计算未充分利用图像的空间关联性。针对上述问题,在基于窗口划分的视觉Transformer方法上提出一种超像素引导的策略,其可以自适应地选择相关窗口进行全局交互。首先,设计基于窗口交互的Top-N交叉注意力机制(TNCA),动态选择与目标图像窗口最相似的前N个窗口,并在通道维度上聚合图像窗口的信息,充分考虑图像非局部自相似性;其次,通过超像素分割引导的方式,显著提升窗口内局部特征的表达力,同时在通道维度上增强空间特征的关联性;最后,构建一个层次化的自适应交互超像素引导的Transformer去噪网络(AISGFormer)。实验结果表明,AISGFormer在SIDD和DND真实图像数据集上的峰值信噪比(PSNR)分别为39.98 dB和40.06 dB,与其他先进网络相比分别提升了0.02 dB~14.33 dB和0.02 dB~7.63 dB,AISGFormer更能交互局部与全局的信息和细节,自适应地利用自相似性来抑制区域相似噪声。

关键词: 低光图像去噪, Transformer, 交叉注意力, 非局部自相似性, 真实图像噪声, 超像素

CLC Number:

TP391.41

SONG Quanzhen, CHEN Zuojun, QIN Pinle, ZENG Jianchao. Superpixel Guide for Transformer Low-Light Image Denoising Method[J]. Computer Engineering, 2026, 52(2): 186-196.

宋泉臻, 陈作钧, 秦品乐, 曾建潮. 基于超像素引导的Transformer低光图像去噪方法[J]. 计算机工程, 2026, 52(2): 186-196.

/ Recommend / Download Citations

URL: https://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0070426

https://www.ecice06.com/EN/Y2026/V52/I2/186

References

[1] 陈钱. 先进夜视成像技术发展探讨[J]. 红外与激光工程,2022, 51(2): 9-16. CHEN Q. Discussions on the development of advanced night vision imaging technology[J]. Infrared and Laser Engineering, 2022, 51(2): 9-16. (in Chinese)
[2] GUO P Y, ASIF M S, MA Z. Low-light color imaging via cross-camera synthesis[J]. IEEE Journal of Selected Topics in Signal Processing, 2022, 16(4): 828-842.
[3] FENG H S, WANG L Z, WANG Y Z, et al. Learnability enhancement for low-light raw image denoising: a data perspective[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(1): 370-387.
[4] YINGKUN H, JUN X, MINGXIA L, et al. NLH: a blind pixel-level non-local method for real-world image denoising[J]. IEEE Transactions on Image Processing, 2020, 29: 5121-5135.
[5] YANG W, WANG W, HUANG H, et al. Sparse gradient regularized deep retinex network for robust low-light image enhancement[J]. IEEE Transactions on Image Processing, 2021, 30: 2072-2086.
[6] HUANG H F, YANG W H, HU Y Y, et al. Towards low light enhancement with RAW images[J]. IEEE Transactions on Image Processing, 2022, 31: 1391-1405.
[7] CUI Y, KNOLL A. PSNet: towards efficient image restoration with self-attention[J]. IEEE Robotics and Automation Letters, 2023, 8(9): 5735-5742.
[8] HASSANI A, WALTON S, LI J C, et al. Neighborhood attention transformer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2023: 6185-6194.
[9] XIAO Y, YUAN Q Q, JIANG K, et al. TTST: a top-k token selective transformer for remote sensing image super-resolution[J]. IEEE Transactions on Image Processing, 2024, 33: 738-752.
[10] TU Z Z, TALEBI H, ZHANG H, et al. MAXIM: multi-axis MLP for image processing[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2022: 5759-5770.
[11] ZHANG S S, MENG N, LAM E Y. LRT: an efficient low-light restoration transformer for dark light field images[J]. IEEE Transactions on Image Processing, 2023, 32: 4314-4326.
[12] 朱凯, 李理, 张彤, 等. 基于Transformer的多阶段运动模糊图像修复网络[J]. 计算机工程, 2024, 50(9): 276-285. ZHU K, LI L, ZHANG T, et al. Multi-stage motion blur image restoration network based on Transformer[J]. Computer Engineering, 2024, 50(9): 276-285. (in Chinese)
[13] ANTONI B, BARTOMEU C, JEAN M M. A non-local algorithm for image denoising[J]. Computer Vision and Pattern Recognition, 2005, 2: 60-65.
[14] DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. IEEE Transactions on Image Processing, 2007, 16(8): 2080-2095.
[15] ZHANG K, ZUO W M, CHEN Y J, et al. Beyond a Gaussian denoiser: residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3142-3155.
[16] GUO S, YAN Z F, ZHANG K, et al. Toward convolutional blind denoising of real photographs[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2019: 1712-1722.
[17] 高煜宝, 文志诚. 基于注意力机制的双路解码器图像去噪方法[J]. 计算机工程, 2024, 50(9): 324-332. GAO Y B, WEN Z C. Dual decoder image denoising method based on attention mechanism[J]. Computer Engineering, 2024, 50(9): 324-332. (in Chinese)
[18] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[EB/OL].[2024-07-05]. https://arxiv.org/abs/2010.11929.
[19] 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 (CVPR). Washington D.C.,USA:IEEE Press,2021: 12294-12305.
[20] WANG Z, CUN X, BAO J, et al. Uformer: a general U-shaped transformer for image restoration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2022: 17662-17672.
[21] JIALE Z, YULUN Z, JINJIN G, et al. Xformer: hybrid X-shaped transformer for image denoising[EB/OL].[2024-07-05]. https://arxiv.org/abs/2303.06440.
[22] 刘凯, 任洪逸, 李蓥,等. 基于交叉模态注意力特征增强的医学视觉问答[J]. 计算机工程, 2025, 51(6): 49-56. LIU K, REN H Y, LI Y, et al. Medical visual question answering based on cross-modal attention feature enhancement[J]. Computer Engineering, 2025, 51(6): 49-56. (in Chinese)
[23] ACHANTA R, SHAJI A, SMITH K, et al. SLIC superpixels compared to state-of-the-art superpixel methods[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(11): 2274-2282.
[24] LIU Y J, YU C C, YU M J, et al. Manifold SLIC: a fast method to compute content-sensitive superpixels[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2016: 651-659.
[25] ACHANTA R, SUSSTRUNK S. Superpixels and polygons using simple non-iterative clustering[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2017: 4895-4904.
[26] ZHANG A, REN W, LIU Y, et al. Lightweight image super-resolution with superpixel token interaction[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Washington D.C.,USA:IEEE Press,2023: 12682-12691.
[27] PAN Y J, WEN C, ZHAO X L, et al. Irregular tensor representation for superpixel- guided hyperspectral image denoising[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 12-23.
[28] ZHOU M, XU Z, TONG R K. Superpixel-guided class-level denoising for unsupervised domain adaptive fundus image segmentation without source data[J]. Computers in Biology and Medicine, 2023, 162: 107061.
[29] ABDELHAMED A, LIN S, BROWN M S. A high-quality denoising dataset for smartphone cameras[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2018: 1692-1700.
[30] PLOTZ T, ROTH S. Benchmarking denoising algorithms with real photographs[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2017: 2750-2759.
[31] DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. IEEE Transactions on Image Processing, 2007, 16(8): 2080-2095.
[32] PARAS M, ZHU L, NING X, et al. Improving extreme low-light image denoising via residual learning[C]//Proceedings of the IEEE International Conference on Multimedia and Exposition. Washington D.C.,USA:IEEE Press,2019: 916-921.
[33] ZAMIR S W, ARORA A, KHAN S, et al. CycleISP: real image restoration via improved data synthesis[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2020: 2693-2702.
[34] ZAMIR S W, ARORA A, KHAN S, et al. Learning enriched features for real image restoration and enhancement[EB/OL].[2024-07-05]. https://arxiv.org/abs/2003.06792.
[35] REN C, HE X H, WANG C C, et al. Adaptive consistency prior based deep network for image denoising[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C.,USA:IEEE Press,2021: 8592-8602.

Please choose a citation manager

Content to export