In order to address the limitations of existing low-light image enhancement (LLIE) algorithms, such as inadequate utilization of frequency-domain information and insufficient cross-domain interaction between frequency and spatial domains, a LLIE algorithm was proposed based on Fourier convolution and wavelet transform self-attention mechanism, termed F-WNet. The core innovation of the algorithm lied in combining Fourier convolution and wavelet transform self-attention to construct a dual-branch network structure, achieving complementary enhancement of frequency-domain and spatial-domain information at the neck of the encoder-decoder framework. The Fourier convolution branch effectively captured global frequency-domain features through fast Fourier convolution residual blocks. The wavelet transform self-attention branch optimized the self-attention mechanism using wavelet transform, which significantly reduced computational complexity while establishing long-range dependencies, thereby enhancing the modeling capability for local details and regional relationships. The outputs of the two branches were adaptively fused through a channel attention fusion module with dynamic weight allocation, further improving feature representation. Finally, the proposed method was evaluated on seven publicly available LLIE benchmark datasets. Experimental results show that the proposed algorithm effectively enhances the brightness of low-light images, reliably suppresses noise, robustly maintains color accuracy, and demonstrates significant advantages in preserving texture details. F-WNet achieves the best performance on 13 evaluation metrics. On the three paired datasets, it attains peak signal-to-noise ratio (PSNR) of 26.977, 27.794, and 28.592 dB, structural similarity (SSIM) of 0.881, 0.897, and 0.943, and learned perceptual image patch similarity (LPIPS) scores of 0.096, 0.120, and 0.052, respectively. On three unpaired datasets, blind/referenceless image spatial quality evaluator (BRISQUE) scores of 25.84, 20.54, and 22.09 and natural image quality evaluator (NIQE) values of 3.75, 4.15, and 3.91 are achieved, respectively. The proposed feasible Fourier-wavelet joint enhancement algorithm effectively utilizes frequency-domain information, establishing a novel cross-domain interaction mechanism that offers a feasible new solution for low-light image enhancement.
TianBowen, DingJianwei, HuZirui, et al. Low-light image enhancement and denoising algorithm integrating zero-reference deep curve[J]. Science Technology and Engineering, 2025, 25(2): 704-712.
[8]
WeiC, WangW, YangW, et al. Deep Retinex decomposition for low-light enhancement[C]// Proceedings of the British Machine Vision Conference. Newcastle: BMVA Press, 2018: 1-12.
[9]
ZhangY H, ZhangJ, GuoX. Kindling the darkness: a practical low-light image enhancer[C]// Proceedings of the 27th ACM International Conference on Multimedia. New York: ACM, 2019: 1632-1640.
HuZirui, DingJianwei, TianBowen. Retinex low-light image enhancement algorithm based on residual and multi-scale attention mechanism[J]. Science Technology and Engineering, 2024, 24(35): 15118-15125.
[12]
GuoC L, LiC Y, GuoJ C, et al. Zero-reference deep curve estimation for low-light image enhancement[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1777-1786.
[13]
JiangY, GongX, LiuD, et al. EnlightenGAN: deep light enhancement without paired supervision[J]. IEEE Transactions on Image Processing, 2021, 30: 2340-2349.
ChenBin, ChenKeyuan, WuShiqian. Low-light image enhancement algorithm based on cross-domain feature fusion[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2437002.
[18]
XuX G, WangR X, FuC W, et al. SNR-Aware low-light image enhancement[C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 17693-17703.
[19]
CaiY H, BianH, LinJ, et al. Retinexformer: one-stage Retinex-based transformer for low-light image enhancement[C]// Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision. Paris: IEEE, 2023: 12470-12479.
[20]
YanQ S, FengY X, ZhangC, et al. You only need one color space: an efficient network for low-light image enhancement[J]. IEEE Transactions on Image Processing, 2024, 33: 1234-1245.
[21]
YuH, ZhengN S, ZhouM, et al. Frequency and spatial dual guidance for image dehazing[M]. Cham: Springer, 2022: 181-198.
[22]
HuangJ, LiuY, ZhaoF, et al. Deep Fourier-based exposure correction network with spatial-frequency interaction[M]. Cham: Springer, 2022: 163-180.
[23]
WangC, WuH, JinZ. FourLLIE: boosting low-light image enhancement by Fourier frequency information[C]// Proceedings of the 31st ACM International Conference on Multimedia. Ottawa: ACM, 2023: 7459-7469.
[24]
TanJ H, PeiS, QinW, et al. Wavelet-based Mamba with Fourier adjustment for low-light image enhancement[M]. Singapore: Springer, 2025: 160-175.
[25]
ChiL, JiangB, MuY. Fast Fourier convolution[C]// Proceedings of the 34th International Conference on Neural Information Processing Systems. Vancouver: Curran Associates, Inc., 2020: 4479-4488.
[26]
VaswaniA, ShazeerN, ParmarN, et al. Attention is all you need[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates, 2017: 6000-6010.
[27]
HuJ, ShenL, SunG. Squeeze-and-excitation networks[C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 7132-7141.
[28]
WangZ, BovikA C, SheikhH R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13(4): 600-612.
[29]
JohnsonJ, AlahiA, LiF F. Perceptual losses for real-time style transfer and super-resolution[C]// Computer Vision-ECCV 2016: 14th European Conference. Berlin: Springer, 2016: 694-711.
[30]
SeifG, AndroutsosD. Edge-based loss function for single image super-resolution[C]// Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing. Calgary: IEEE, 2018: 1468-1472.
[31]
YangW, WangW, HuangH, et al. Sparse gradient regularized deep Retinex network for robust low-light image enhancement[J]. IEEE Transactions on Image Processing, 2021, 30: 2072-2086.
[32]
KingmaD P, BaJ. Adam: a method for stochastic optimization[C]// Proceedings of the 3rd International Conference on Learning Representations. San Diego: ICLR, 2015: 1-15.
[33]
LoshchilovI, HutterF. SGDR: stochastic gradient descent with warm restarts[C]// Proceedings of the 5th International Conference on Learning Representations. Toulon: OpenReview. net, 2017: 3983.
[34]
ZamirS W, AroraA, KhanS, et al. Learning enriched features for real image restoration and enhancement[C]// Computer Vision-ECCV 2020: 16th European Conference. Glasgow, UK: Springer, 2020: 492-511.
[35]
LiuR S, MaL, ZhangJ A, et al. Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhancement[C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 10556-10565.
[36]
WangY, WanR, YangW, et al. Low-light image enhancement with normalizing flow[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2022: 2604-2612.
[37]
ZamirS W, AroraA, KhanS, et al. Restormer: efficient Transformer for high-resolution image restoration[C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA: IEEE, 2022: 5718-5729.
[38]
WangT, ZhangK H, ShenT R, et al. Ultra-high-definition low-light image enhancement: a benchmark and transformer-based method[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2023: 2689-2697.
[39]
LiC, GuoC, LoyC C. Learning to enhance low-light image via zero-reference deep curve estimation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(8): 4225-4238.
[40]
WangZ D, CunX D, BaoJ M, et al. Uformer: a general U-shaped transformer for image restoration[C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 17662-17672.
[41]
ZhangR, IsolaP, EfrosA A, et al. The unreasonable effectiveness of deep features as a perceptual metric[C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT: IEEE, 2018: 586-595.
[42]
ZengH, CaiJ, LiL, et al. Learning image-adaptive 3D lookup tables for high performance photo enhancement in real-time[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(4): 2058-2073.
[43]
YangW, WangS, FangY, et al. From fidelity to perceptual quality: a semi-supervised approach for low-light image enhancement[C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA: IEEE, 2020: 3060-3069.
[44]
ZhouS, LiC, ChangeL C. LEDNet: joint low-light enhancement and deblurring in the dark[C]// Computer Vision-ECCV 2022. Cham: Springer, 2022: 573-589.
[45]
FuZ Q, YangY, TuX T, et al. Learning a simple low-light image enhancer from paired low-light instances[C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC: IEEE, 2023: 22252-22261.
[46]
ZhangJ, WangX, LiY, et al. PRNet: low-light image enhancement based on Fourier transform[J]. IEEE Transactions on Instrumentation and Measurement, 2025, 74: 1-14.