Understanding no-reference low-light metrics
Optional · supplementary note 9 of 14 for DLCV 09C · Low-Light Image Enhancement. Nothing in the main text depends on it; citation numbers refer to the reference list at the bottom of this page.
Navigate: this page and this track: Navigate: this page and this track
On this page
- CLAHE: equalizing locally without amplifying noise
- More deep LLIE methods
- Counting resolution, arithmetic and memory
- A bilateral-grid experiment
- Implementing and checking trilinear LUTs
- Fitting and slicing a bilateral grid in detail
- Learnable LUT variants in detail
- Structural re-parameterization step by step
- Understanding no-reference low-light metrics
- Efficient network building blocks
- Banding, halos and temporal stability
- Bilateral grids and color-table mathematics
- From retouching to low light
- A lightness-order exercise
- KinD: repair reflectance while controlling illumination
- Retinex-Net: decompose, adjust and denoise
- EnlightenGAN: unpaired appearance learning with content constraints
- Zero-DCE and Zero-DCE++: learn curves without reference images
- RUAS: search the priors inside an unrolled enhancement model
- SCI: train with calibration, deploy one illumination block
- Retinexformer: illumination guides restoration
- IAT: local affine maps and a global camera-style transform
- Multinex: fuse analytic priors into a Retinex residual
- CPGA-Net: channel priors, gamma estimation and learned fusion
- CPGA-Net+: attention, BDSF and stronger variants
- HDRNet: deep bilateral learning for real-time image enhancement
- Image-adaptive 3D LUTs (Zeng et al., TPAMI 2022)
Deep Learning for Computer Vision
Plain version. Some measures target what goes wrong in enhancement specifically: Did the method swap which regions are brighter than others? Are the colors vivid or washed out? How much noise did brightening reveal?
Lightness Order Error (LOE). The NPE paper [13] argues that a natural enhancement should keep the relative order of lightness in the scene: if a lamp was brighter than a wall before, it should still be after, because that order carries the direction of the light and the sense of depth. LOE counts how often the order of two pixels flips. With the lightness of the input and that of the enhanced image,
where is the number of pixels, if and 0 otherwise, and is exclusive-or, so is the number of pixels whose order relative to changed. Lower is better. Three practical points:
- The cost is , so implementations evaluate LOE on a subsampled image. The raw value grows with , so numbers are comparable only at the same subsampling. Many papers do not state theirs.
- The reference is the input, not a clean image. Any monotonic tone curve applied in the max channel gets LOE , including doing nothing at all. And because the noise in the dark input already scrambles the order of nearby pixels, a method that removes noise is penalized.
- LOE says nothing about whether the result is bright enough. It is a guard against unnatural local tone mapping, not a quality score.
Colorfulness. Hasler and Süsstrunk [62] fit a simple colorfulness measure to perceptual ratings, using opponent-color axes on 8-bit RGB:
where and are the mean and standard deviation over the image, and the weight 0.3 comes from the authors’ fit. Higher means more colorful. It is useful as a descriptive statistic for desaturation (luma-only enhancement, Figure 2) or oversaturation, but it has no notion of the correct color, and noise in the chroma channels inflates and .
Noise level of the output. Since brightening reveals noise (Figure 1), it helps to report the noise of the output. Immerkær’s fast estimator [63] filters the image with a mask that cancels smooth structure and keeps noise,
where is the image size and the sum runs over interior pixels. It assumes white Gaussian noise, so it overestimates on textured images and underestimates spatially correlated noise, which is what real camera pipelines and many networks produce. Use it to compare outputs of the same input, not across scenes.
The snippet below recreates the dark coffee image of Figure 1 and compares five “enhancements” plus the clean photo.
Code · Python · 53 lines
import numpy as np
from scipy import ndimage as ndi
from skimage import color, data, exposure, img_as_float
rng = np.random.default_rng(0)
clean = img_as_float(data.coffee())
def synth_low_light(x, k=0.05, peak=2000.0, read_std=8e-4, gamma=2.2): # as in Figure 1
lin = k * x ** gamma
lin = rng.poisson(lin * peak) / peak
lin = lin + rng.normal(0, read_std, lin.shape)
return np.round(np.clip(lin, 0, 1) ** (1 / gamma) * 255) / 255
def loe(ref, out, step=8):
"""Lightness Order Error on a strided subsample (m = 50 x 75 = 3750 pixels here)."""
L, Le = (im.max(axis=2)[::step, ::step].ravel() for im in (ref, out))
U, Ue = L[:, None] >= L[None, :], Le[:, None] >= Le[None, :]
return (U ^ Ue).sum(axis=1).mean() # mean over x of RD(x)
def colorfulness(im):
"""Hasler-Suesstrunk colorfulness, RGB scaled to 0..255."""
R, G, B = (255 * im[..., c] for c in range(3))
rg, yb = R - G, 0.5 * (R + G) - B
return np.hypot(rg.std(), yb.std()) + 0.3 * np.hypot(rg.mean(), yb.mean())
def noise_sigma(im):
"""Immerkaer's noise estimate on BT.601 luma, in 8-bit units."""
Y = 255 * (0.299 * im[..., 0] + 0.587 * im[..., 1] + 0.114 * im[..., 2])
N = np.array([[1, -2, 1], [-2, 4, -2], [1, -2, 1]])
H, W = Y.shape
return np.sqrt(np.pi / 2) / (6 * (W - 2) * (H - 2)) * np.abs(ndi.convolve(Y, N)[1:-1, 1:-1]).sum()
dark = synth_low_light(clean)
lab = color.rgb2lab(dark); L = lab[..., 0] / 100
he = lab.copy(); he[..., 0] = 100 * exposure.equalize_hist(L)
cl = lab.copy(); cl[..., 0] = 100 * exposure.equalize_adapthist(L, clip_limit=0.01)
outputs = {
"dark input": dark,
"gain x20 (linear)": np.clip(20 * dark ** 2.2, 0, 1) ** (1 / 2.2),
"gamma 0.4": dark ** 0.4,
"HE on L*": np.clip(color.lab2rgb(he), 0, 1),
"CLAHE on L*": np.clip(color.lab2rgb(cl), 0, 1),
"clean reference": clean,
}
for name, im in outputs.items():
print(f"{name:18s} mean {im.mean():.2f} LOE {loe(dark, im):6.1f} "
f"colorfulness {colorfulness(im):5.1f} noise {noise_sigma(im):5.2f}")
# dark input mean 0.10 LOE 0.0 colorfulness 23.1 noise 3.32
# gain x20 (linear) mean 0.39 LOE 0.3 colorfulness 90.0 noise 12.93
# gamma 0.4 mean 0.34 LOE 0.0 colorfulness 61.6 noise 11.19
# HE on L* mean 0.48 LOE 334.7 colorfulness 28.3 noise 16.17
# CLAHE on L* mean 0.40 LOE 657.6 colorfulness 27.1 noise 23.06
# clean reference mean 0.39 LOE 322.0 colorfulness 76.9 noise 4.69
Read the numbers together, not one at a time:
- Brightness alone ranks the wrong way. HE has the highest mean (0.48) but also a large LOE, the second-highest noise and almost no color left. CLAHE’s local remapping breaks lightness order the most.
- Colorfulness is fooled by noise. The linear gain scores 90, higher than the clean photo’s 77, partly because amplified chroma noise raises and .
- LOE penalizes restoration. The clean photo, the ideal answer, gets LOE 322 against the noisy input, while doing nothing at all gets 0. Low LOE is necessary for a natural tone curve, not sufficient for quality.
- Only the noise estimate separates the clean photo (4.7) from the bright but noisy outputs (11 to 23). No single NR number captures “well exposed, natural color, and clean”.
Learned no-reference assessors
Neural and vision-language assessors (NIMA, MUSIQ, CLIP-IQA, MANIQA, Q-Align and others) are described in DLCV 09 and DLCV 09; they can all be applied to LLIE outputs and share common implementations, e.g. the IQA-PyTorch toolbox [81]. Three points are specific to low light:
- Resolution. Avoiding resizing matters for LLIE, where noise and fine texture are exactly what resizing would hide, so an assessor that takes the image at its native resolution, such as MUSIQ [82], is a natural fit.
- Exposure prompts and independence. CLIP-IQA [83] can be probed with an exposure-specific prompt pair such as “Bright photo.” / “Dark photo.”, which is convenient for exposure, but the exact wording changes the score. The same CLIP-similarity idea is used as a training signal by CLIP-LIT [42], so a CLIP-based metric should not be the only judge of a CLIP-trained enhancer.
- LLIE-specific assessors. Kannan et al. [84] (2023) collected human opinion scores for images restored by different LLIE methods, benchmarked existing NR-IQA models on them, and proposed an opinion-unaware model that learns distortion-aware features of restored low-light images by self-supervised contrastive learning. Datasets of this kind are the right test bed for deciding which metric to trust on enhancement outputs.
Caveats for low light. Learned assessors are trained mostly on daylight photos. They tend to rate brighter, sharper and more saturated images higher, which favors over-enhancement, and a deliberately dark night scene can be penalized even when it is the faithful result. Their scores depend on the checkpoint (which training set), the input size and the implementation, so report those. And when a method is tuned with a learned metric or the same backbone in its loss, that metric is no longer an independent judge.
User studies and paired comparisons
Mean opinion scores (standardized in ITU-R BT.500 [85]), Bradley–Terry paired comparison [86] and good practice for running a study are described in DLCV 09. Two low-light papers show the pattern:
- Zero-DCE [18] had 15 subjects rate results on a 1–5 scale over the 202 images of NPE, LIME, MEF, DICM and VV, with instructions to look for over- or under-exposure, color deviation, and unnatural texture or noise.
- EnlightenGAN [17] used paired comparisons: 9 subjects compared pairs of outputs from five methods on 23 images, judging visible noise, exposure artifacts and unrealistic color or texture, and a Bradley–Terry fit gave the ranking.
The criteria given to the raters (exposure, color deviation, noise, unnatural texture) are a good starting list for a new low-light study.
Choosing and reporting no-reference metrics
The general comparison of NR metrics (family, training requirement, direction, sensitivity) is the metric-by-metric summary table in the DLCV 09 notes, and its recommended practice applies unchanged. The table below keeps only what changes for low-light outputs.
| Metric | Better | Caveat for low light |
|---|---|---|
| NIQE [58] | Lower | Nearly blind to global exposure and color; can prefer noisy, high-contrast outputs |
| BRISQUE [59] | Lower | Trained on daylight distortions; ignores color cast |
| PI [60], [61] | Lower | Designed for super-resolution; rewards detail, including amplified noise |
| LOE [13] (task-specific) | Lower | Depends on subsampling; 0 for identity; penalizes denoising |
| Colorfulness [62] (task-specific, descriptive) | Context-dependent | No notion of correct color; inflated by chroma noise |
| Noise estimate [63] (task-specific, descriptive) | Lower | Assumes white noise; confused by texture and correlated noise |
| NIMA [87] | Higher | Aesthetic models reward attractive, not faithful, results |
| MUSIQ [82] | Higher | Trained mostly on daylight photos; checkpoint-dependent |
| CLIP-IQA [83] | Higher | Prompt wording matters; not independent of CLIP-trained enhancers |
| MANIQA [88] | Higher | Not trained on exposure errors specifically |
| Q-Align [89] | Higher | Heavy; biases of the base model; version-dependent |
| MOS / paired comparison [85], [86] | Higher MOS / higher strength | Costly; small studies are noisy; protocol must be reported |
What to do on top of the general practice:
- On paired sets, lead with full-reference metrics (PSNR, SSIM, LPIPS) under the standard protocol, and add NR scores only as a complement.
- On unpaired sets, report several NR metrics from different families (one NSS, one or two learned, and LOE), state the implementation, checkpoint and image size, and show many visual examples, including failures.
- Expect disagreement. NSS and learned metrics often rank methods differently, and both can disagree with people; the survey of Li et al. [15] notes the frequent mismatch between metric scores and visual quality, and Liu and Fan [66] find a further gap between image quality and downstream task performance. When metrics disagree, say so instead of picking the favorable one.
References
- Shyandram, “簡介Deep Learning Low-Light Image Enhancement (LLIE)” (Introduction to deep-learning low-light image enhancement, in Chinese), Medium post, 2024; updated 2026. link
- A. Foi, M. Trimeche, V. Katkovnik and K. Egiazarian, “Practical Poissonian-Gaussian noise modeling and fitting for single-image raw-data,” IEEE Transactions on Image Processing, vol. 17, no. 10, pp. 1737–1754, 2008. doi
- K. He, J. Sun and X. Tang, “Single image haze removal using dark channel prior,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 12, pp. 2341–2353, 2011. doi
- X. Guo, Y. Li and H. Ling, “LIME: Low-light image enhancement via illumination map estimation,” IEEE Transactions on Image Processing, vol. 26, no. 2, pp. 982–993, 2017. doi
- ITU-R, “Recommendation BT.601-7: Studio encoding parameters of digital television for standard 4:3 and wide-screen 16:9 aspect ratios,” 2011. ITU
- Q. Yan, Y. Feng, C. Zhang, et al., “HVI: A new color space for low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2025. arXiv
- S. M. Pizer, E. P. Amburn, J. D. Austin, et al., “Adaptive histogram equalization and its variations,” Computer Vision, Graphics, and Image Processing, vol. 39, no. 3, pp. 355–368, 1987. doi
- K. Zuiderveld, “Contrast limited adaptive histogram equalization,” in Graphics Gems IV, Academic Press, 1994, pp. 474–485. doi
- E. H. Land, “The retinex theory of color vision,” Scientific American, vol. 237, no. 6, pp. 108–128, 1977. doi
- D. J. Jobson, Z. Rahman and G. A. Woodell, “Properties and performance of a center/surround retinex,” IEEE Transactions on Image Processing, vol. 6, pp. 451–462, 1997. doi
- D. J. Jobson, Z. Rahman and G. A. Woodell, “A multiscale retinex for bridging the gap between color images and the human observation of scenes,” IEEE Transactions on Image Processing, vol. 6, no. 7, pp. 965–976, 1997. doi
- X. Dong, Y. Pang and J. Wen, “Fast efficient algorithm for enhancement of low lighting video,” ACM SIGGRAPH 2010 Posters, 2010. SIGGRAPH history doi
- S. Wang, J. Zheng, H.-M. Hu and B. Li, “Naturalness preserved enhancement algorithm for non-uniform illumination images,” IEEE Transactions on Image Processing, vol. 22, no. 9, pp. 3538–3548, 2013. doi
- C. Lee, C. Lee and C.-S. Kim, “Contrast enhancement based on layered difference representation of 2D histograms,” IEEE Transactions on Image Processing, vol. 22, no. 12, pp. 5372–5384, 2013. doi
- C. Li, C. Guo, L. Han, J. Jiang, M.-M. Cheng, J. Gu and C. C. Loy, “Low-light image and video enhancement using deep learning: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 12, pp. 9396–9416, 2022. arXiv
- C. Wei, W. Wang, W. Yang and J. Liu, “Deep retinex decomposition for low-light enhancement,” in Proc. British Machine Vision Conference (BMVC), 2018. arXiv project
- Y. Jiang, X. Gong, D. Liu, et al., “EnlightenGAN: Deep light enhancement without paired supervision,” IEEE Transactions on Image Processing, vol. 30, pp. 2340–2349, 2021. arXiv
- C. Guo, C. Li, J. Guo, et al., “Zero-reference deep curve estimation for low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2020, pp. 1777–1786. arXiv
- C. Li, C. Guo and C. C. Loy, “Learning to enhance low-light image via zero-reference deep curve estimation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 8, 2022. arXiv
- R. Liu, L. Ma, J. Zhang, X. Fan and Z. Luo, “Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2021, pp. 10556–10565. arXiv
- L. Ma, T. Ma, R. Liu, X. Fan and Z. Luo, “Toward fast, flexible, and robust low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2022, pp. 5627–5636. arXiv
- Y. Cai, H. Bian, J. Lin, H. Wang, R. Timofte and Y. Zhang, “Retinexformer: One-stage Retinex-based transformer for low-light image enhancement,” in Proc. IEEE/CVF ICCV, 2023. arXiv code
- Y. Wang, R. Wan, W. Yang, H. Li, L.-P. Chau and A. C. Kot, “Low-light image enhancement with normalizing flow,” in Proc. AAAI Conference on Artificial Intelligence, vol. 36, no. 3, pp. 2604–2612, 2022. arXiv
- X. Yi, H. Xu, H. Zhang, L. Tang and J. Ma, “Diff-Retinex: Rethinking low-light image enhancement with a generative diffusion model,” in Proc. IEEE/CVF ICCV, 2023. arXiv
- J. Hou, Z. Zhu, J. Hou, H. Liu, H. Zeng and H. Yuan, “Global structure-aware diffusion process for low-light image enhancement,” in Advances in Neural Information Processing Systems (NeurIPS), 2023. arXiv
- H. Jiang, A. Luo, X. Liu, S. Han and S. Liu, “LightenDiffusion: Unsupervised low-light image enhancement with latent-Retinex diffusion models,” in Proc. European Conference on Computer Vision (ECCV), 2024. arXiv
- C. He, C. Fang, Y. Zhang, et al., “Reti-Diff: Illumination degradation image restoration with Retinex-based latent diffusion model,” in Proc. International Conference on Learning Representations (ICLR), 2025. proceedings
- M. Gharbi, J. Chen, J. T. Barron, S. W. Hasinoff and F. Durand, “Deep bilateral learning for real-time image enhancement,” ACM Transactions on Graphics (Proc. SIGGRAPH), vol. 36, no. 4, art. 118, 2017. doi arXiv
- H. Zeng, J. Cai, L. Li, Z. Cao and L. Zhang, “Learning image-adaptive 3D lookup tables for high performance photo enhancement in real-time,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 4, 2022. doi arXiv
- K. G. Lore, A. Akintayo and S. Sarkar, “LLNet: A deep autoencoder approach to natural low-light image enhancement,” Pattern Recognition, vol. 61, pp. 650–662, 2017. doi
- L. Tao, C. Zhu, J. Song, T. Lu, H. Jia and X. Xie, “Low-light image enhancement using CNN and bright channel prior,” in Proc. IEEE International Conference on Image Processing (ICIP), 2017, pp. 3215–3219. doi
- S. W. Zamir, A. Arora, S. Khan, et al., “Learning enriched features for real image restoration and enhancement,” in Proc. European Conference on Computer Vision (ECCV), 2020, pp. 492–511. arXiv
- X. Xu, R. Wang, C.-W. Fu and J. Jia, “SNR-aware low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2022, pp. 17693–17703. doi
- Z. Cui, K. Li, L. Gu, S. Su, P. Gao, Z. Jiang, Y. Qiao and T. Harada, “You only need 90K parameters to adapt light: A light weight transformer for image enhancement and exposure correction,” in Proc. British Machine Vision Conference (BMVC), 2022. paper arXiv
- X. Ding, X. Zhang, N. Ma, J. Han, G. Ding and J. Sun, “RepVGG: Making VGG-style ConvNets great again,” in Proc. IEEE/CVF CVPR, 2021, pp. 13728–13737. doi arXiv
- Y. Zhang, J. Zhang and X. Guo, “Kindling the darkness: A practical low-light image enhancer,” in Proc. ACM International Conference on Multimedia (ACM MM), 2019, pp. 1632–1640. doi
- Y. Zhang, X. Guo, J. Ma, W. Liu and J. Zhang, “Beyond brightening low-light images,” International Journal of Computer Vision, vol. 129, no. 4, pp. 1013–1037, 2021. doi
- J. Bai, Y. Yin, Q. He, Y. Li and X. Zhang, “RetinexMamba: Retinex-based Mamba for low-light image enhancement,” arXiv:2405.03349, 2024. arXiv
- S.-E. Weng, S.-G. Miaou and R. Christanto, “A lightweight low-light image enhancement network via channel prior and gamma correction,” International Journal of Pattern Recognition and Artificial Intelligence, vol. 39, no. 12, art. 2554013, 2025. doi code manuscript
- S.-E. Weng, C.-Y. Hsiao, L.-W. Lu, et al., “Rethinking theoretical illumination for efficient low-light image enhancement” (CPGA-Net+), arXiv:2409.05274, 2024. arXiv
- W. Wu, J. Weng, P. Zhang, X. Wang, W. Yang and J. Jiang, “URetinex-Net: Retinex-based deep unfolding network for low-light image enhancement,” in Proc. IEEE/CVF CVPR, 2022, pp. 5891–5900. doi
- Z. Liang, C. Li, S. Zhou, R. Feng and C. C. Loy, “Iterative prompt learning for unsupervised backlit image enhancement,” in Proc. IEEE/CVF ICCV, 2023. arXiv
- C. Yang, M. Jin, X. Jia, Y. Xu and Y. Chen, “AdaInt: Learning adaptive intervals for 3D lookup tables on real-time image enhancement,” in Proc. IEEE/CVF CVPR, 2022, pp. 17522–17531. CVF arXiv
- C. Yang, M. Jin, Y. Xu, R. Zhang, Y. Chen and H. Liu, “SepLUT: Separable image-adaptive lookup tables for real-time image enhancement,” in Proc. European Conference on Computer Vision (ECCV), 2022, pp. 201–217. doi arXiv
- T. Wang, Y. Li, J. Peng, Y. Ma, X. Wang, F. Song and Y. Yan, “Real-time image enhancer via learnable spatial-aware 3D lookup tables,” in Proc. IEEE/CVF ICCV, 2021, pp. 2471–2480. CVF arXiv
- C. Liu, H. Yang, J. Fu and X. Qian, “4D LUT: Learnable context-aware 4D lookup table for image enhancement,” IEEE Transactions on Image Processing, vol. 32, pp. 4742–4756, 2023. doi arXiv
- W. Kim and N. I. Cho, “Image-adaptive 3D lookup tables for real-time image enhancement with bilateral grids,” in Proc. European Conference on Computer Vision (ECCV), 2024, pp. 91–108. doi ECCV
- C. Chen, Q. Chen, J. Xu and V. Koltun, “Learning to see in the dark,” in Proc. IEEE/CVF CVPR, 2018, pp. 3291–3300. arXiv
- M. V. Conde, G. Geigle and R. Timofte, “InstructIR: High-quality image restoration following human instructions,” in Proc. European Conference on Computer Vision (ECCV), 2024. arXiv
- Z. Cui, G.-J. Qi, L. Gu, S. You, Z. Zhang and T. Harada, “Multitask AET with orthogonal tangent regularity for dark object detection,” in Proc. IEEE/CVF ICCV, 2021. arXiv
- K. A. Hashmi, G. Kallempudi, D. Stricker and M. Z. Afzal, “FeatEnHancer: Enhancing hierarchical features for object detection and beyond under low-light vision,” in Proc. IEEE/CVF ICCV, 2023. arXiv
- M. Dehghani, A. Arnab, L. Beyer, A. Vaswani and Y. Tay, “The efficiency misnomer,” in Proc. International Conference on Learning Representations (ICLR), 2022. ICLR arXiv
- A. Ignatov, R. Timofte, S. Liu, et al., “Learned smartphone ISP on mobile GPUs with deep learning, Mobile AI & AIM 2022 challenge: Report,” in Proc. European Conference on Computer Vision (ECCV) Workshops, 2023, pp. 44–70. doi arXiv
- Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600–612, 2004. doi
- R. Zhang, P. Isola, A. A. Efros, E. Shechtman and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proc. IEEE/CVF CVPR, 2018. arXiv
- K. Ma, K. Zeng and Z. Wang, “Perceptual quality assessment for multi-exposure image fusion,” IEEE Transactions on Image Processing, vol. 24, no. 11, pp. 3345–3356, 2015. doi
- V. Vonikakis, “Datasets” (source of the VV test images), personal web page. link
- A. Mittal, R. Soundararajan and A. C. Bovik, “Making a ‘completely blind’ image quality analyzer,” IEEE Signal Processing Letters, vol. 20, no. 3, pp. 209–212, 2013. doi
- A. Mittal, A. K. Moorthy and A. C. Bovik, “No-reference image quality assessment in the spatial domain,” IEEE Transactions on Image Processing, vol. 21, no. 12, pp. 4695–4708, 2012. doi
- Y. Blau, R. Mechrez, R. Timofte, T. Michaeli and L. Zelnik-Manor, “The 2018 PIRM challenge on perceptual image super-resolution,” in Proc. European Conference on Computer Vision (ECCV) Workshops, 2018. arXiv
- C. Ma, C.-Y. Yang, X. Yang and M.-H. Yang, “Learning a no-reference quality metric for single-image super-resolution,” Computer Vision and Image Understanding, vol. 158, pp. 1–16, 2017. doi arXiv
- D. Hasler and S. E. Süsstrunk, “Measuring colorfulness in natural images,” in Proc. SPIE 5007, Human Vision and Electronic Imaging VIII, 2003. doi
- J. Immerkær, “Fast noise variance estimation,” Computer Vision and Image Understanding, vol. 64, no. 2, pp. 300–302, 1996. doi
- J. Liao, S. Hao, R. Hong and M. Wang, “GT-Mean loss: A simple yet effective solution for brightness mismatch in low-light image enhancement,” in Proc. IEEE/CVF ICCV, 2025. arXiv
- Y. P. Loh and C. S. Chan, “Getting to know low-light images with the Exclusively Dark dataset,” Computer Vision and Image Understanding, vol. 178, pp. 30–42, 2019. arXiv
- F. Liu and L. Fan, “A review of advancements in low-light image enhancement using deep learning,” Neurocomputing, vol. 652, art. 131052, 2025. arXiv
- J. Liu, D. Xu, W. Yang, M. Fan and H. Huang, “Benchmarking low-light image enhancement and beyond,” International Journal of Computer Vision, vol. 129, no. 4, pp. 1153–1184, 2021. doi project
- W. Yang, W. Wang, H. Huang, S. Wang and J. Liu, “Sparse gradient regularized deep Retinex network for robust low-light image enhancement,” IEEE Transactions on Image Processing, vol. 30, pp. 2072–2086, 2021. doi
- J. Hai, Z. Xuan, R. Yang, et al., “R2RNet: Low-light image enhancement via Real-low to Real-normal Network,” Journal of Visual Communication and Image Representation, vol. 90, art. 103712, 2023. arXiv
- R. Wang, X. Xu, C.-W. Fu, et al., “Seeing dynamic scene in the dark: A high-quality video dataset with mechatronic alignment,” in Proc. IEEE/CVF ICCV, 2021, pp. 9680–9689. doi code
- V. Bychkovsky, S. Paris, E. Chan and F. Durand, “Learning photographic global tonal adjustment with a database of input/output image pairs,” in Proc. IEEE CVPR, 2011, pp. 97–104. dataset
- J. Cai, S. Gu and L. Zhang, “Learning a deep single image contrast enhancer from multi-exposure images,” IEEE Transactions on Image Processing, vol. 27, no. 4, pp. 2049–2062, 2018. doi
- X. Liu, Z. Wu, A. Li, et al., “NTIRE 2024 challenge on low light image enhancement: Methods and results,” in Proc. IEEE/CVF CVPR Workshops, 2024. arXiv
- X. Liu, Z. Wu, F.-A. Vasluianu, et al., “NTIRE 2025 challenge on low light image enhancement: Methods and results,” in Proc. IEEE/CVF CVPR Workshops, 2025. arXiv
- S. Zheng, Y. Ma, J. Pan, C. Lu and G. Gupta, “Low-light image and video enhancement: A comprehensive survey and beyond,” arXiv:2212.10772, 2022. arXiv
- E. Adhikarla, Y. Liu and B. D. Davison, “Diffusion models for low-light image enhancement: A multi-perspective taxonomy and performance analysis,” arXiv:2510.05976, 2025. arXiv
- J. Yan, C. Tu, W. Zhang, Z. Wang, P. Cao, Q. Lin, Y. Fang, X. Liu, Z. Wu, Z. Zhou and R. Timofte, “NTIRE 2026 challenge on efficient low light image enhancement: Methods and results,” NTIRE 2026 workshop report, arXiv:2605.02212, 2026. arXiv
- O. Ronneberger, P. Fischer and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in Proc. Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2015, pp. 234–241. doi arXiv
- J. Chen, A. Adams, N. Wadhwa and S. W. Hasinoff, “Bilateral guided upsampling,” ACM Transactions on Graphics (Proc. SIGGRAPH Asia), vol. 35, no. 6, 2016. doi project
- K. He, J. Sun and X. Tang, “Guided image filtering,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 6, pp. 1397–1409, 2013. doi
- C. Chen and J. Mo, “IQA-PyTorch: PyTorch toolbox for image quality assessment,” software (pip package
pyiqa). GitHub - J. Ke, Q. Wang, Y. Wang, P. Milanfar and F. Yang, “MUSIQ: Multi-scale image quality transformer,” in Proc. IEEE/CVF ICCV, 2021. arXiv
- J. Wang, K. C. K. Chan and C. C. Loy, “Exploring CLIP for assessing the look and feel of images,” in Proc. AAAI Conference on Artificial Intelligence, 2023. arXiv
- V. Kannan, S. Malik, N. C. Babu and R. Soundararajan, “Quality assessment of low-light restored images: A subjective study and an unsupervised model,” IEEE Access, vol. 11, pp. 68216–68230, 2023. doi arXiv
- ITU-R, “Recommendation BT.500-15: Methodologies for the subjective assessment of the quality of television images,” 2023. ITU
- R. A. Bradley and M. E. Terry, “Rank analysis of incomplete block designs: The method of paired comparisons,” Biometrika, vol. 39, no. 3–4, pp. 324–345, 1952. doi
- H. Talebi and P. Milanfar, “NIMA: Neural image assessment,” IEEE Transactions on Image Processing, 2018. arXiv
- S. Yang, T. Wu, S. Shi, et al., “MANIQA: Multi-dimension attention network for no-reference image quality assessment,” in Proc. IEEE/CVF CVPR Workshops, 2022, pp. 1190–1199. doi arXiv
- H. Wu, Z. Zhang, W. Zhang, C. Chen, et al., “Q-Align: Teaching LMMs for visual scoring via discrete text-defined levels,” in Proc. International Conference on Machine Learning (ICML), PMLR 235, 2024. arXiv
- A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto and H. Adam, “MobileNets: Efficient convolutional neural networks for mobile vision applications,” arXiv:1704.04861, 2017. arXiv
- M. Sandler, A. Howard, M. Zhu, A. Zhmoginov and L.-C. Chen, “MobileNetV2: Inverted residuals and linear bottlenecks,” in Proc. IEEE/CVF CVPR, 2018, pp. 4510–4520. doi arXiv
- W.-S. Lai, J.-B. Huang, O. Wang, E. Shechtman, E. Yumer and M.-H. Yang, “Learning blind video temporal consistency,” in Proc. European Conference on Computer Vision (ECCV), 2018, pp. 179–195. doi arXiv
- S. W. Hasinoff, D. Sharlet, R. Geiss, A. Adams, J. T. Barron, F. Kainz, J. Chen and M. Levoy, “Burst photography for high dynamic range and low-light imaging on mobile cameras,” ACM Transactions on Graphics (Proc. SIGGRAPH Asia), vol. 35, no. 6, 2016. doi
- J. Chen, S. Paris and F. Durand, “Real-time edge-aware image processing with the bilateral grid,” ACM Transactions on Graphics (Proc. SIGGRAPH), vol. 26, no. 3, art. 103, 2007. doi
- C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images,” in Proc. IEEE International Conference on Computer Vision (ICCV), 1998, pp. 839–846. doi
- S. Paris and F. Durand, “A fast approximation of the bilateral filter using a signal processing approach,” International Journal of Computer Vision, vol. 81, no. 1, pp. 24–52, 2009. doi
- H. Wu, S. Zheng, J. Zhang and K. Huang, “Fast end-to-end trainable guided filter,” in Proc. IEEE/CVF CVPR, 2018, pp. 1838–1847. doi arXiv
- R. Wang, Q. Zhang, C.-W. Fu, X. Shen, W.-S. Zheng and J. Jia, “Underexposed photo enhancement using deep illumination estimation,” in Proc. IEEE/CVF CVPR, 2019, pp. 6842–6850. doi CVF
- Y. Lin, Z. Fu, K. Wen, T. Ye, S. Chen, G. Meng, Y. Wang, et al., “DPLUT: Unsupervised low-light image enhancement with lookup tables and diffusion priors,” in Proc. AAAI Conference on Artificial Intelligence, vol. 39, no. 5, pp. 5316–5324, 2025. doi arXiv
- K. Dabov, A. Foi, V. Katkovnik and K. Egiazarian, “Image denoising by sparse 3-D transform-domain collaborative filtering,” IEEE Transactions on Image Processing, 2007. doi
- K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” ICLR, 2015. arXiv
- J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li and L. Fei-Fei, “ImageNet: A large-scale hierarchical image database,” CVPR, 2009. doi
- A. Brateanu, T. Mu, C. O. Ancuti and C. Ancuti, “Multinex: Lightweight Low-light Image Enhancement via Multi-prior Retinex,” in Proc. IEEE/CVF CVPR, 2026. paper supplement