An Adaptive Portrait Image Enhancement Algorithm Based on Facial and Scene Statistics in Complex Lighting

Guozhi Hu; Robiatul A’dawiah Jamaluddin

doi:10.54097/hywwks72

Authors

Guozhi Hu 1 Kuala Lumpur University of Science and Technology (KLUST), Kuala Lumpur, Malaysia; 2 Beijing Kingsha Technology Co., Ltd., Beijing, China
Robiatul A’dawiah Jamaluddin Kuala Lumpur University of Science and Technology (KLUST), Kuala Lumpur, Malaysia

DOI:

https://doi.org/10.54097/hywwks72

Keywords:

Color normalization, complex lighting, highlight suppression, portrait enhancement, statistical profiling

Abstract

Portrait image quality in user-provided photographs is frequently compromised by highly variable capture conditions, including underexposure, facial shadows cast by side lighting, and color casts introduced by indoor illuminants. This paper presents an interpretable, fully ablatable workflow for portrait-perception-based color normalization. The algorithm builds a multidimensional statistical profile of the face and the surrounding scene in the LAB and HSV color spaces, and uses this profile to drive adaptive color-cast correction with a dead-zone-and-damping mechanism, localized skin-tone equalization through large-scale Gaussian diffusion, nonlinear luminance mapping with highlight suppression, and detail-aware texture restoration. By separating luminance from chrominance, the method avoids the color crosstalk typical of RGB-space adjustments. On a dataset of 120 portraits covering 12 typical complex-lighting scenarios, our method raises the central L-mean from 144.64 to 165.55 while constraining the overall highlight ratio to 0.72%. In the overexposed group, the method suppresses the highlight ratio to 1.12%, avoiding the failure state (8.57%) produced by conventional brightness enhancement, and in the low-light group it lifts the central L-mean from 128.37 to 170.05. Ablation experiments confirm that the color-normalization module is the dominant contributor to these gains. The workflow offers a robust, reproducible, and interpretable alternative to black-box enhancement filters for portrait normalization under complex lighting.

Downloads

Download data is not yet available.

References

[1] Land, E. H. (1977). The retinex theory of color vision. Scientific American, 237(6), 108–128. https://doi.org/10.1038/scientificamerican1277-108

[2] Jobson, D. J., Rahman, Z., & Woodell, G. A. (1997). A multiscale retinex for bridging the gap between color images and the human observation of scenes. IEEE Transactions on Image Processing, 6(7), 965–976. https://doi.org/10.1109/83.597272

[3] Pizer, S. M., et al. (1987). Adaptive histogram equalization and its variations. Computer Vision, Graphics, and Image Processing, 39(3), 355–368. https://doi.org/10.1016/0734-189X(87)90156-4

[4] Zuiderveld, K. (1994). Contrast limited adaptive histogram equalization. In P. S. Heckbert (Ed.), Graphics gems IV (pp. 474–485). Academic Press.

[5] Guo, X., Li, Y., & Ling, H. (2017). LIME: Low-light image enhancement via illumination map estimation. IEEE Transactions on Image Processing, 26(2), 982–993. https://doi.org/10.1109/TIP.2016.2631944

[6] Wei, C., Wang, W., Yang, W., & Liu, J. (2018). Deep retinex decomposition for low-light enhancement. In Proceedings of the British Machine Vision Conference (BMVC). https://doi.org/10.5244/C.32.127

[7] Guo, C., et al. (2020). Zero-reference deep curve estimation for low-light image enhancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 1780–1789). https://doi.org/10.1109/CVPR42600.2020.01791

[8] Jiang, Y., et al. (2021). EnlightenGAN: Deep light enhancement without paired supervision. IEEE Transactions on Image Processing, 30, 2340–2349. https://doi.org/10.1109/TIP.2021.3052733

[9] Wu, W., et al. (2022). URetinex-Net: Retinex-based deep unfolding network for low-light image enhancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 5901–5910). https://doi.org/10.1109/CVPR52688.2022.00584

[10] Buchsbaum, G. (1980). A spatial processor model for object colour perception. Journal of the Franklin Institute, 310(1), 1–26. https://doi.org/10.1016/0016-0032(80)90033-8

[11] Ebner, M. (2007). Color constancy. John Wiley & Sons.

[12] van de Weijer, J., Gevers, T., & Gijsenij, A. (2007). Edge-based color constancy. IEEE Transactions on Image Processing, 16(9), 2207–2214. https://doi.org/10.1109/TIP.2007.903229

[13] Viola, P., & Jones, M. (2001). Rapid object detection using a boosted cascade of simple features. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (Vol. 1, pp. 511–518). https://doi.org/10.1109/CVPR.2001.990517

[14] Rudin, L. I., Osher, S., & Fatemi, E. (1992). Nonlinear total variation based noise removal algorithms. Physica D: Nonlinear Phenomena, 60(1–4), 259–268. https://doi.org/10.1016/0167-2789(92)90242-F

[15] Zheng, P., et al. (2024). Bilateral reference for high-resolution dichotomous image segmentation. CAAI Artificial Intelligence Research, 3, 9150038. https://doi.org/10.1016/j.caair.2024.9150038

[16] Polesel, A., Ramponi, G., & Mathews, V. J. (2000). Image enhancement via adaptive unsharp masking. IEEE Transactions on Image Processing, 9(3), 505–510. https://doi.org/10.1109/83.827816