<p>Most existing low-light image enhancement methods suffer from excess noise from inevitably noise in images captured under low-light conditions and amplified noise during the enhancement process, which affects the enhancement effect and subsequent downstream visual tasks. To balance the enhancement effect and noise suppression, we propose a zero-reference Low-light Image Enhancement and Denoising network utilizing implicit and explicit priors, named Zero-LIED. Specifically, the Image Decomposition and Denoising Network (IDDN) decomposes the input image into illumination and reflectance components based on Retinex theory, and applies deep image prior to achieve implicit denoising during training. The Illumination Adjustment Network (IAN) achieves pixel-wise exposure adjustment of the illumination component by estimating gamma correction parameter maps, rather than using only one parameter for the entire image. The IDDN and IAN training process is guided by five loss functions as explicit priors. The final enhanced image is obtained by fusing the reflectance component decomposed by IDDN and the illumination component enhanced by IAN. Extensive quantitative and qualitative experiments have shown that the proposed Zero-LIED achieves enhancement effects comparable to those of state-of-the-art methods with less data and simpler network structure.</p>