RetinexGuI: Retinex-Guided Iterative Illumination Estimation Method for Low Light Images

TL;DR

RetinexGuI introduces a lightweight Retinex-guided LLIE framework that iteratively refines illumination in the log domain while keeping reflectance fixed, achieving $\mathcal{O}(N)$ computational complexity. By operating on the HSV V channel and using a cascaded refinement with gamma correction on S, it delivers natural color reproduction and real-time enhancement without training data. A key result is the closed-form convergence $I_l^{(\infty)} = \frac{(I_l^{(0)})^2}{I_l^{(0)} - 1}$ that explains stable performance with few iterations. Experiments on VE-LOL-L, RELLISUR, and LoLi-Street show competitive PSNR/SSIM and superior runtime compared to several baselines, highlighting practical viability and potential for integration with learning-based systems. The work also points to future directions in robustness under extreme illumination and code release for broader adoption.

Abstract

In recent years, there has been a growing interest in low-light image enhancement (LLIE) due to its importance for critical downstream tasks. Current Retinex-based methods and learning-based approaches have shown significant LLIE performance. However, computational complexity and dependencies on large training datasets often limit their applicability in real-time applications. We introduce RetinexGuI, a novel and effective Retinex-guided LLIE framework to overcome these limitations. The proposed method first separates the input image into illumination and reflection layers, and iteratively refines the illumination while keeping the reflectance component unchanged. With its simplified formulation and computational complexity of $\mathcal{O}(N)$, our RetinexGuI demonstrates impressive enhancement performance across three public datasets, indicating strong potential for large-scale applications. Furthermore, it opens promising directions for theoretical analysis and integration with deep learning approaches. The source code will be made publicly available at https://github.com/etuspars/RetinexGuI once the paper is accepted.

Figures (2)

• Figure 1: Visual comparison of low-light image enhancement results produced by different methods on three representative images from the three different datasets.
• Figure 2: Visual comparison of using different numbers of iterations.