Researchers Publish New Low-Light Image Enhancement Model in Scientific Reports

On July 31, 2026, researchers Kerui Wu, Nan Di, and Jipeng Huang published a new low-light image enhancement model in Scientific Reports. The model utilizes the original Retinex multiplicative decomposition model and an objective function with variational characteristics to preserve texture details. By solving two convex subproblems alternately using the Alternating Direction Method of Multipliers (ADMM), the method achieves superior performance compared to mainstream advanced algorithms across three public datasets.

Low-light image enhancement methodology

• ▪Kerui Wu, Nan Di, and Jipeng Huang published a low-light image enhancement model in Scientific Reports on July 31, 2026.
• ▪The low-light image enhancement research was supported by the Department of Science, Technology and Information of the Ministry of Education under project number 8091B042240.

Retinex multiplicative decomposition model

• ▪Converting Retinex models into additive decomposition models using logarithmic transformations can lead to the loss of texture details in the reflectance map.
• ▪The proposed model adopts the original Retinex multiplicative decomposition model rather than converting it into an additive decomposition model.

ADMM optimization algorithm

• ▪The proposed model constructs an objective function with variational characteristics that is decomposed into two convex subproblems.
• ▪The two convex subproblems of the objective function are alternately solved using the Alternating Direction Method of Multipliers.

Texture detail preservation

• ▪A weight matrix based on spatial domain statistical features is introduced in both structural and texture regularization terms to enhance texture information.
• ▪Experimental results demonstrate that the generated reflectance map retains richer texture details and achieves effective enhancement of low-light images.

Public dataset benchmark performance

• ▪Quantitative and qualitative comparisons with several mainstream advanced algorithms demonstrate that the proposed method achieves superior performance in low-light image enhancement tasks.
• ▪The proposed low-light image enhancement method was evaluated and tested on three commonly used public datasets.

1 source