HistRetinex: Optimizing Retinex model in Histogram Domain for Efficient Low-Light Image Enhancement

TL;DR

HistRetinex redefines Retinex-based low-light enhancement in the histogram domain to achieve real-time performance on large images. By introducing histogram location $H_B(\cdot)$ and histogram count $H_C(\cdot)$ matrices, it models the interaction between illumination and reflectance in histogram space and solves a two-level optimization to estimate their histograms, followed by histogram-based enhancement and matching. The framework decouples complexity from image resolution, yielding a per-iteration cost tied to the grayscale level count $l$ and a total complexity of $O(I\cdot l^2)$, with reported runtimes around $1.86$ seconds for $1000\times664$ images and substantial time savings over existing methods. Empirical results on multiple datasets show improved visibility and metrics (PSNR, SSIM, NIQE, LOE) compared to state-of-the-art Retinex and histogram methods, indicating strong practical impact for fast, high-quality low-light image enhancement. The approach preserves Retinex principles while enabling scalable, real-time performance suitable for real-world vision tasks.

Abstract

Retinex-based low-light image enhancement methods are widely used due to their excellent performance. However, most of them are time-consuming for large-sized images. This paper extends the Retinex model from the spatial domain to the histogram domain, and proposes a novel histogram-based Retinex model for fast low-light image enhancement, named HistRetinex. Firstly, we define the histogram location matrix and the histogram count matrix, which establish the relationship among histograms of the illumination, reflectance and the low-light image. Secondly, based on the prior information and the histogram-based Retinex model, we construct a novel two-level optimization model. Through solving the optimization model, we give the iterative formulas of the illumination histogram and the reflectance histogram, respectively. Finally, we enhance the low-light image through matching its histogram with the one provided by HistRetinex. Experimental results demonstrate that the HistRetinex outperforms existing enhancement methods in both visibility and performance metrics, while executing 1.86 seconds on 1000*664 resolution images, achieving a minimum time saving of 6.67 seconds.

Figures (7)

• Figure 1: Execution time of existing low-light enhancement algorithms in low-light images with different resolutions.
• Figure 2: The framework of the histogram-based Retinex model.
• Figure 3: Histograms of various components in the histogram-based Retinex model. (a) original image; (b) illumination component; (c) reflectance component; (d) estimate V channel component; (e) original V channel component.
• Figure 4: The performance metrics of these algorithms in low-light images. (a) PSNR; (b) SSIM.
• Figure 5: Enhancement results of these algorithms in low-light images. (a) #Candle. (b) #BelgiumHouse; (c) #House; (d) #Daybreak&Nightfall (11); (e) #night (52); (f) #Parking. *From top to bottom, each row denotes low-light image, LIME, STAR, AHPCE, WVM, SSD, RRM, BRAIN, LR3M, FSTAR and proposed algorithm, respectively.