Enhancement of Low-Light Images through Illumination Estimation and Local Steering Kernel

: Low-light image enhancement remains a crucial challenge in the domain of computer vision and multimedia applications. Images captured under insufficient lighting conditions often suffer from poor visibility, reduced contrast, noise amplification, and color distortion, which can significantly degrade the performance of downstream visual processing tasks. Although several enhancement techniques have been proposed to restore visibility and simulate normal lighting conditions, these methods often introduce artifacts such as overexposed regions, amplified noise, and unnatural color tones. To address these limitations, we propose a novel low-light image enhancement method that leverages Retinex theory for illumination component estimation, combined with a local steering kernel for denoising and color correction. The illumination component is separated from the reflectance to accurately simulate the natural lighting environment, while the local steering kernel adaptively smoothens and corrects colors, preserving important structural details. The proposed technique is tested on a diverse set of low light images captured under various environmental settings. Both qualitative and quantitative evaluations demonstrate that our method significantly improves image clarity, enhances contrast, and reduces noise without introducing visible artifacts. Comparative analysis with existing state-of-the-art algorithms highlights the superior performance of our approach in preserving naturalness and structural fidelity. This makes our solution particularly suitable for enhancing images in surveillance, photography, and autonomous systems operating under low light scenarios.

Image processing; Enhancement; Retinex; Steering kernel; Low-light image

1. A. Agarwal and V. Dhar, “Big Data, AI, and Machine Learning in Financial Services,” J. Financial Perspectives, vol. 2, no. 3, pp. 1–14, 2014.
2. I. Jeong and C. Lee, “An optimization-based approach to gamma correction parameter estimation for low-light image enhancement,” Multimedia Tools Appl., vol. 80, pp. 18027–18042, 2021.
3. M. Li, J. Liu, W. Yang, X. Sun, and Z. Guo, “Structure-revealing low-light image enhancement via robust Retinex model,” IEEE Trans. Image Process., vol. 27, no. 6, pp. 2828–2841, Jun. 2018.
4. Q. Mu, X. Wang, Y. Wei, and Z. Li, “Low and non-uniform illumination color image enhancement using weighted guided image filtering,” Comput. Vis. Media, vol. 7, pp. 529–546, 2021.
5. H. Tang, H. Zhu, H. Tao, and C. Xie, “An improved algorithm for low-light image enhancement based on RetinexNet,” Appl. Sci., vol. 12, no. 14, p. 7268, 2022.
6. Q. Zhang, H. Tang, Y. Li, B. Han, and J. Li, “Improved method based on Retinex and Gabor for the surface defect enhancement of aluminum strips,” Metals, vol. 13, no. 1, p. 118, 2023.