Transmittance-Guided Structure-Texture Decomposition for Nighttime Image Dehazing

Francesco Moretti; Giulia Bianchi; Andrea Gallo

cs.CVMar 31, 2026

Transmittance-Guided Structure-Texture Decomposition for Nighttime Image Dehazing

Authors: Francesco Moretti, Giulia Bianchi, Andrea Gallo

Organizations: Maharaja Agrasen University

Abstract

Nighttime images captured under hazy conditions suffer from severe quality degradation, including low visibility, color distortion, and reduced contrast, caused by the combined effects of atmospheric scattering, absorption by suspended particles, and non-uniform illumination from artificial light sources. While existing nighttime dehazing methods have achieved partial success, they typically address only a subset of these issues, such as glow suppression or brightness enhancement, without jointly tackling the full spectrum of degradation factors. In this paper, we propose a two-stage nighttime image dehazing framework that integrates transmittance correction with structure-texture layered optimization. In the first stage, we introduce a novel transmittance correction method that establishes boundary-constrained initial transmittance maps and subsequently applies region-adaptive compensation and normalization based on whether image regions correspond to light source areas. A quadratic Gaussian filtering scheme operating in the YUV color space is employed to estimate the spatially varying atmospheric light map. The corrected transmittance map and atmospheric light map are then used in conjunction with an improved nighttime imaging model to produce the initial dehazed image. In the second stage, we propose a STAR-YUV decomposition model that separates the dehazed image into structure and texture layers within the YUV color space. Gamma correction and MSRCR-based color restoration are applied to the structure layer for illumination compensation and color bias correction, while Laplacian-of-Gaussian filtering is applied to the texture layer for detail enhancement. A novel two-phase fusion strategy, comprising nonlinear Retinex-based fusion of the enhanced layers followed by linear blending with the initial dehazing result, yields the final output.

Figures & tables

Method	Image1	Image2	Image3	Average
DCP	20.506	37.320	15.277	24.368
DehazeNet	19.221	34.211	15.783	23.072
CEEF	26.025	29.485	25.481	26.997
GMLC	17.690	35.502	15.661	22.951
Fb	19.849	20.973	17.238	19.353
IAT	21.684	20.519	16.871	19.691

Table 1: CIEDE2000 color difference evaluation (lower is better). Bold and underlined values indicate the best and second-best results, respectively.

Method	Detail	Color	Natural.	Visual	Overall
DCP	3.2	3.4	3.6	3.3	3.4
DehazeNet	3.0	3.2	3.5	3.2	3.2
CEEF	3.6	4.0	3.8	3.7	3.8
GMLC	4.2	3.2	4.0	3.9	4.2
Fb	3.1	4.1	3.8	4.0	3.8
IAT	3.0	4.0	3.9	3.9	3.7

Table 2: Subjective evaluation scores (1–5 scale) by 30 human assessors. Bold values indicate the best results per column.

	ZS330 Samples			HC770 Samples
Method	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$
DCP	3.789	6.535	4.310	14.053	0.611	4.292	6.659	2.461
DehazeNet	3.122	5.596	4.782	13.422	0.441	3.613	5.740	2.692
CEEF	6.763	7.057	4.296	15.920	0.669	7.076	7.176	2.205
GMLC	7.548	6.990	4.380	15.570	0.398	7.814	6.936	2.464
Fb	3.347	6.934	4.344	14.290	0.612	4.335	7.100	2.515

Table 3: Quantitative comparison on representative images from ZS330 (no-reference only) and HC770 (all metrics).

\uparrow

: higher is better;

\downarrow

: lower is better. Bold and underlined values indicate best and second-best results.

	ZS330			HC770
Method	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$
DCP	4.526	6.764	2.779	9.899	0.556	3.968	6.835	2.775
DehazeNet	3.807	6.100	3.127	12.105	0.592	3.229	6.663	3.054
CEEF	6.503	7.042	2.600	11.469	0.618	6.592	7.078	2.721
GMLC	7.020	8.321	2.689	14.096	0.460	7.083	6.676	2.640
Fb	3.235	6.795	3.136	13.536	0.693	2.590	6.684	3.392

Table 4: Quantitative comparison on the full ZS330 and HC770 test sets.

\uparrow

: higher is better;

\downarrow

: lower is better. Bold and underlined values indicate best and second-best results.

	ZS330			HC770
Method	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	AG $\uparrow$	IE $\uparrow$	NIQE $\downarrow$
w/o T	19.065	16.289	5.314	12.959	0.380	5.046	7.701	7.607
w/o Dehaze	7.304	7.500	4.226	15.547	0.592	6.330	7.501	4.357
w/o STAR	5.723	7.113	3.748	16.387	0.779	5.834	7.394	3.711
Full Model	7.836	7.461	2.683	17.024	0.765	7.371	7.595	2.702

Table 5: Ablation study results on the ZS330 and HC770 test sets. Bold and underlined values indicate best and second-best results.

Method	480 $\times$ 360	550 $\times$ 733	1000 $\times$ 653	Average
DCP	8.54	21.04	33.52	21.03
DehazeNet	1.16	3.71	4.25	3.04
CEEF	0.64	2.36	2.55	1.85
GMLC	6.08	13.90	40.74	20.24
Fb	5.51	17.49	27.56	16.85
IAT	8.85	19.24	32.63	20.24

Table 6: Runtime comparison (in seconds) across different image resolutions. Bold values indicate the fastest method.

Image Dehazing

Image dehazing restores clear scene details and colors from images obscured by haze, fog, or smoke.

See all 35 papers

Apr 4, 2026

HistoFusionNet: Histogram-Guided Fusion and Frequency-Adaptive Refinement for Nighttime Image Dehazing

Nighttime image dehazing remains a challenging low-level vision problem due to the joint presence of haze, glow, non-uniform illumination, color distortion, and sensor noise, which often invalidate assumptions commonly used in daytime dehazing. To address these challenges, we propose HistoFusionNet, a transformer-enhanced architecture tailored for nighttime image dehazing by combining histogram-guided representation learning with frequency-adaptive feature refinement. Built upon a multi-scale encoder-decoder backbone, our method introduces histogram transformer blocks that model long-range dependencies by grouping features according to their dynamic-range characteristics, enabling more effective aggregation of similarly degraded regions under complex nighttime lighting. To further improve restoration fidelity, we incorporate a frequency-aware refinement branch that adaptively exploits complementary low- and high-frequency cues, helping recover scene structures, suppress artifacts, and enhance local details. This design yields a unified framework that is particularly well suited to the heterogeneous degradations encountered in real nighttime hazy scenes. Extensive experiments and highly competitive performance of our method on the NTIRE 2026 Nighttime Image Dehazing Challenge benchmark demonstrate the effectiveness of the proposed method. Our team ranked 1st among 22 participating teams, highlighting the robustness and competitive performance of HistoFusionNet. The code is available at: https://github.com/heydarimo/Night-Time-Dehazing

Mohammad Heydari, Wei Dong, Shahram Shirani +2

McMaster University

Image Dehazing Image Restoration

Ask PDF

Aug 11, 2026

HNDiff: Haze-Noise Diffusion for Image Dehazing

Existing diffusion-based methods have recently made significant progress in image dehazing. However, they typically neglect the physics of haze formation and reconstruct clean images from pure Gaussian noise, thereby limiting their restoration potential. To address this issue, we propose Haze-Noise Diffusion (HNDiff), a novel diffusion framework that embeds the atmospheric scattering model as an inductive bias. By grounding diffusion in physical principles, HNDiff ensures that the restoration aligns more closely with underlying mechanisms of haze formation. In its forward process, we introduce joint haze-noise diffusion with a haze-aware noise scheduler, which progressively adds both haze and noise to an image. Essentially, the scheduler adapts noise levels according to haze density, meaning that regions with heavier haze receive stronger noise injection to encourage content generation, while clearer regions receive lighter noise to better preserve details, which directly links the forward degradation process with the physics of haze. In the reverse process, we then derive a physically consistent dehazing-denoising process that simultaneously removes haze and noise to restore a clean image in a manner aligned with the forward degradation process. To further enhance practicality, we propose Latent HNDiff, which compiles clean latent priors that can be seamlessly integrated into existing dehazing networks to boost performance. Extensive experiments show that our work significantly improves leading dehazing backbones and achieves state-of-the-art results on benchmark datasets. The project page is available at https://jin-ting-he.github.io/HNDiff .

Jin-Ting He, Fu-Jen Tsai, Yan-Tsung Peng +3

Department of Computer Science, National Yang Ming Chiao Tung University, Taiwan · National Tsing Hua University, Taiwan · National Chengchi University, Taiwan +1

Physics-Informed Diffusion Models Diffusion Models for Image Restoration Physics-Informed ML Image Dehazing

Ask PDF

Apr 7, 2026

CLIP-Guided Data Augmentation for Night-Time Image Dehazing

Nighttime image dehazing faces a more complex degradation pattern than its daytime counterpart, as haze scattering couples with low illumination, non-uniform lighting, and strong light interference. Under limited supervision, this complexity aggravates domain drift and training instability, since target-domain samples are scarce while naively introducing external data may weaken adaptation due to distribution mismatch. This paper presents our solution to the NTIRE 2026 Night Time Image Dehazing Challenge, built as a unified framework that integrates domain-aligned data construction, stage-wise training, and inference-time enhancement. Specifically, a pre-trained CLIP visual encoder screens candidate external samples by similarity to construct training data closer to the target domain. NAFNet is then trained in two stages, first adapting to the target domain and then expanding to broader degradation patterns. At inference time, TLC, x8 self-ensemble, and weighted snapshot fusion are combined to improve output stability. Rather than relying on complex network redesign, the proposed framework offers a practical and effective pipeline for nighttime image dehazing.

Xining Ge, Weijun Yuan, Gengjia Chang +2

Hangzhou Dianzi University · Jinan University · Hefei University of Technology +2

Image Dehazing Data Augmentation

Ask PDF

Abstract

Image Dehazing

Explore similar work

HistoFusionNet: Histogram-Guided Fusion and Frequency-Adaptive Refinement for Nighttime Image Dehazing

HNDiff: Haze-Noise Diffusion for Image Dehazing

CLIP-Guided Data Augmentation for Night-Time Image Dehazing