Contact Citation Download Result
Abstract Framework Highlights
Home

Haze Hue and Haze Saturation Priors for Single Image Dehazing

Sobhan Kanti Dhara, Mayukh Roy, Debashis Sen

Use the slider to compare input and dehazed output.

TestImage1 TestImage2
TestImage1 TestImage2
TestImage1 TestImage2
Abstracticon

Abstract

Due to scattering and absorption by suspended particles in hazy atmospheres, images captured suffer from visibility attenuation, low contrast and color cast issues. Handling a wide variety of hazy conditions with such degradations to reduce haze in images is a challenging task. In this regard, our paper proposes a novel image dehazing framework based on new hazy image priors, namely, the Haze Hue Prior and the Haze Saturation Prior. The Haze Hue Prior (HHP) presents distinguishing properties of hue in non-color-cast and color-cast hazy images, which is exploited to classify hazy images into these two categories. The Haze Saturation Prior (HSP) provides characteristic properties of saturation in hazy conditions, which is utilized to handle color distortion. The proposed framework also leverages a novel foreground-aware strategy to compute the atmospheric light and a depth-aware strategy to estimate the transmission map, which are then used in the standard atmospheric light scattering model to achieve dehazing. Extensive quantitative evaluations demonstrate that our proposed framework outperforms the state-of-the-art across a wide range of natural and synthetic hazy images. Subjective evaluations show that our proposed framework effectively removes color cast, significantly reduces haze, and preserves the natural image appearance without introducing noticeable distortion.

Frameworkicon

Framework

In this paper, we propose an approach based on atmospheric light relevant priors for dehazing hazy images that may contain color cast. To this end, we design novel hazy image priors, namely, haze hue prior (HHP) and haze saturation prior (HSP), which represent haze induced image conditions related to the atmospheric light parameter of Koschmieder's model. These proposed priors are not mere assumptions but represent facts, which are theoretically and empirically proved. In our dehazing approach, HHP is first employed to categorize a hazy image as a color cast or a non-cast one. Then, a color cast hazy image is subjected to a novel color correction procedure based on HSP to diminish the cast. A foreground-aware approach is then proposed to estimate the atmospheric light from the color corrected hazy image or the non-cast hazy image. Transmission values are then computed using the atmospheric light estimate and applying a fast weighted regularized least squares solver on the dark channel of the color corrected or non-cast hazy image. The computed transmission and atmospheric light values are finally used in the Koschmieder's model corresponding to the color corrected or non-cast hazy image to get the restored image. Qualitative and quantitative evaluation using standard datasets and measures show the effectiveness of our approach in classifying hazy images, in diminishing color cast and reducing haze, and in producing visibility-improved natural looking dehazed images. The proposed approach performs well for both real-world and synthetically generated hazy images, and is found to outperform the state-of-the-art.

Framework
Highlightsicon

Highlights

The main novelties and contributions of the paper are:

Resultsicon

Comparison of Results

Real World Hazy Images

Hover on Results for Better View.

Input_1

Input

ALC_1

ALC

FSID_1

FSID

MSBDN_1

MSBDN

DDCP_1

D-DCP

PSD_1

PSD

LDNet_1

LDNet

SGID_1

SGID

D4_1

D4

DeHamer_1

DeHamer

MBTF_1

MB-TF

TOENET_1

TOENet

DEANET_1

DEANet

Ours_1

Ours

Input_2

Input

ALC_2

ALC

FSID_2

FSID

MSBDN_2

MSBDN

DDCP_2

D-DCP

PSD_2

PSD

LDNet_2

LDNet

SGID_2

SGID

D4_2

D4

DeHamer_2

DeHamer

MBTF_2

MB-TF

TOENET_2

TOENet

DEANET_2

DEANet

Ours_2

Ours

Input_3

Input

ALC_3

ALC

FSID_3

FSID

MSBDN_3

MSBDN

DDCP_3

D-DCP

PSD_3

PSD

LDNet_3

LDNet

SGID_3

SGID

D4_3

D4

DeHamer_3

DeHamer

MBTF_3

MB-TF

TOENET_3

TOENet

DEANET_3

DEANet

Ours_3

Ours

Input_4

Input

ALC_4

ALC

FSID_4

FSID

MSBDN_4

MSBDN

DDCP_4

D-DCP

PSD_4

PSD

LDNet_4

LDNet

SGID_4

SGID

D4_4

D4

DeHamer_4

DeHamer

MBTF_4

MB-TF

TOENET_4

TOENet

DEANET_4

DEANet

Ours_4

Ours

Synthetically Generated Hazy Images

Hover on Results for Better View.

Input_synth_1

Input

GT_synth_1

GT

ALC_synth_1

ALC

FSID_synth_1

FSID

MSBDN_synth_1

MSBDN

DDCP_synth_1

D-DCP

PSD_synth_1

PSD

LDNet_synth_1

LDNet

SGID_synth_1

SGID

D4_synth_1

D4

DeHamer_synth_1

DeHamer

MBTF_synth_1

MB-TF

TOENET_synth_1

TOENet

DEANET_synth_1

DEANet

Ours_synth_1

Ours

Input_synth_2

Input

GT_synth_2

GT

ALC_synth_2

ALC

FSID_synth_2

FSID

MSBDN_synth_2

MSBDN

DDCP_synth_2

D-DCP

PSD_synth_2

PSD

LDNet_synth_2

LDNet

SGID_synth_2

SGID

D4_synth_2

D4

DeHamer_synth_2

DeHamer

MBTF_synth_2

MB-TF

TOENET_synth_2

TOENet

DEANET_synth_2

DEANet

Ours_synth_2

Ours

Input_synth_3

Input

GT_synth_3

GT

ALC_synth_3

ALC

FSID_synth_3

FSID

MSBDN_synth_3

MSBDN

DDCP_synth_3

D-DCP

PSD_synth_3

PSD

LDNet_synth_3

LDNet

SGID_synth_3

SGID

D4_synth_3

D4

DeHamer_synth_3

DeHamer

MBTF_synth_3

MB-TF

TOENET_synth_3

TOENet

DEANET_synth_3

DEANet

Ours_synth_3

Ours

Input_synth_4

Input

GT_synth_4

GT

ALC_synth_4

ALC

FSID_synth_4

FSID

MSBDN_synth_4

MSBDN

DDCP_synth_4

D-DCP

PSD_synth_4

PSD

LDNet_synth_4

LDNet

SGID_synth_4

SGID

D4_synth_4

D4

DeHamer_synth_4

DeHamer

MBTF_synth_4

MB-TF

TOENET_synth_4

TOENet

DEANET_synth_4

DEANet

Ours_synth_4

Ours

Downloadsicon

Downloads

Download Paper

Paper

Download Code

Code

Referencesicon

References

[1] Peng, Y., Lu, Z., & Cheng, F., et al. (2019). Image haze removal using airlight white correction, local light filter, and aerial perspective prior. IEEE Trans on Circuits and Systems for Video Technoogy (pp. 1–1). https://doi.org/10.1109/TCSVT.2019.2902795
[2] Kim, S. E., Park, T. H., & Eom, I. K. (2020). Fast single image dehazing using saturation based transmission map estimation. IEEE Trans. Image Process., 29, 1985–1998.
[3] Dong, H., Pan, J., & Xiang, L., et al. (2020). Multi-scale boosted dehazing network with dense feature fusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, (pp. 2157–2167).
[4] Golts, A., Freedman, D., & Elad, M. (2020). Unsupervised single image dehazing using dark channel prior loss. IEEE Trans. Image Process., 29, 2692–2701.
[5] Chen, Z., Wang, Y., & Yang, Y., et al. (2021). PSD: Principled synthetic-to-real dehazing guided by physical priors. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (pp. 7180–7189).
[6] Ullah, H., Muhammad, K., Irfan,M., et al. (2021).Light-dehazenet: A novel lightweight cnn architecture for single image dehazing. IEEE Trans. Image Process., 30, 8968–8982. https://doi.org/10.1109/TIP.2021.3116790
[7] Bai,H., Pan, J.,Xiang,X., et al. (2022). Self-guided image dehazing using progressive feature fusion. IEEE Trans. Image Process., 31, 1217–1229.
[8] Yang,Y.,Wang, C.,&Liu, R., et al (2022) Self-augmented unpaired image dehazing via density and depth decomposition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 2037–2046.
[9] Guo, C.L., Yan, Q., & Anwar, S., et al. (2022). Image dehazing transformer with transmission-aware 3d position embedding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
[10] Qiu, Y., Zhang, K., & Wang, C., et al. (2023). Mb-taylorformer: Multi-branch efficient transformer expanded by taylor formula for image dehazing. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), (pp. 12802–12813).
[11] Yuan Gao, W.X., & Lu, Y. (2023). Let you see in haze and sandstorm: Two-in-one low-visibility enhancement network. IEEE Transactions on Instrumentation and Measurement 72.
[12] Chen, Z., He, Z., & Lu, Z. M. (2024). Dea-net: Single image dehazing based on detail-enhanced convolution and content-guided attention. IEEE Trans. Image Process., 33, 1002–1015.

Citationicon

Citation

Dhara, S.K., Roy, M. & Sen, D. Haze Hue and Haze Saturation Priors for Single Image Dehazing. Int J Comput Vis 134, 116 (2026). https://doi.org/10.1007/s11263-025-02655-5


RIS

TY - JOUR
AU - Dhara, Sobhan K.
AU - Roy, Mayukh
AU - Sen, Debashis
PY - 2026
DA - 2026/02/11
TI - Haze Hue and Haze Saturation Priors for Single Image Dehazing
JO - International Journal of Computer Vision
SP - 116
VL - 134
IS - 3
AB - Due to scattering and absorption by suspended particles in hazy atmospheres, images captured suffer from visibility attenuation, low contrast and color cast issues. Handling a wide variety of hazy conditions with such degradations to reduce haze in images is a challenging task. In this regard, our paper proposes a novel image dehazing framework based on new hazy image priors, namely, the Haze Hue Prior and the Haze Saturation Prior. The Haze Hue Prior (HHP) presents distinguishing properties of hue in non-color-cast and color-cast hazy images, which is exploited to classify hazy images into these two categories. The Haze Saturation Prior (HSP) provides characteristic properties of saturation in hazy conditions, which is utilized to handle color distortion. The proposed framework also leverages a novel foreground-aware strategy to compute the atmospheric light and a depth-aware strategy to estimate the transmission map, which are then used in the standard atmospheric light scattering model to achieve dehazing. Extensive quantitative evaluations demonstrate that our proposed framework outperforms the state-of-the-art across a wide range of natural and synthetic hazy images. Subjective evaluations show that our proposed framework effectively removes color cast, significantly reduces haze, and preserves the natural image appearance without introducing noticeable distortion Project page: https://m14roy.github.io/HHHSPID-Image-Dehazing/
SN - 1573-1405
UR - https://doi.org/10.1007/s11263-025-02655-5
DO - 10.1007/s11263-025-02655-5
ID - Dhara2026
ER -


BIBTeX

@article{article,
author = {Dhara, Sobhan and Roy, Mayukh and Sen, Debashis},
year = {2026},
month = {02},
pages = {},
title = {Haze Hue and Haze Saturation Priors for Single Image Dehazing: Haze Hue
and Haze Saturation Priors for Single Image DehazingS.K. Dhara et al.},
volume = {134},
journal = {International Journal of Computer Vision},
doi = {10.1007/s11263-025-02655-5}
}