To address the issues of prominent noise, color deviation, and low training efficiency in low-light image enhancement, this paper proposes a low-light image enhancement algorithm named ZH-Net, which integrates zero-reference curve and histogram equalization techniques. The algorithm significantly improves the image enhancement effect while increasing the processing speed. Firstly, ZH-Net constructs a lightweight feature extraction network based on a CNN backbone, introduces residual connections to optimize gradient flow, and designs an edge feature enhancement module to better preserve image details. Secondly, the algorithm combines the power function with histogram equalization to greatly enhance the image quality. Experimental results on the LOLv2 dataset demonstrate that the performance of the ZH-Net algorithm surpasses existing mainstream methods, achieving a PSNR of 27.63 dB, SSIM of 0.895, LPIPS reduced to 0.065, with a model parameter count of only 0.07M. The algorithm exhibits significant advantages in restoration accuracy, computational efficiency, and model lightweighting, effectively improving the visual quality of low-light images.