<p>Low-light image enhancement aims to improve brightness, restore details, and suppress artifacts such as Poisson-Gaussian noise in low-light RGB images. Although numerous enhancement models have been proposed, they often struggle with adaptability in complex illumination scenarios, leading to over- or under-enhancement, and rely on fixed parameters that fail to adjust to varying brightness and content. To address these problems, this paper proposes a low-light image enhancement model based on an adaptive enhancement matrix. The proposed algorithm introduces an adaptive weight enhancement module based on deep learning, which dynamically generates a threshold T used to adjust the enhancement matrix. This guides luminance adaptation within an iterative enhancement framework, achieving balanced enhancement while preventing overexposure and insufficient detail restoration. Experiments on benchmark datasets, including LOL-v1, LOL-v2, and LSRW, demonstrate that our method outperforms state-of-the-art techniques by 3.02 dB in PSNR and 8.4% in SSIM on average.</p>