A Dual-Branch Convolutional Neural Network for Low-Light Image Enhancement

Low-light image enhancement is one of the core challenges in computer vision and image processing, aiming to improve the visibility, detail preservation, and color fidelity of images captured under weak illumination. Although traditional methods based on histogram equalization or Retinex theory have made notable progress, their sensitivity to complex noise and tendency toward local overexposure still limit their applicability in real-world scenarios. To address these shortcomings, we propose a lightweight dual-branch convolutional neural network (DualBranch CNN) framework that combines an attention-guided feature enhancement mechanism with a cross-domain noise suppression module. Extensive experiments on two benchmark datasets (ExDark and MIT-Adobe FiveK) demonstrate that our approach outperforms existing algorithms both in objective metrics (PSNR improvement of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 2. 3 \%}$</tex>, SSIM improvement of 8.7 %) and in subjective visual quality, with particularly strong performance in texture detail preservation and highlight suppression. This paper provides a theoretical foundation for deploying low-light enhancement on mobile devices and in realtime video streams, and paves the way for novel techniques in cross-modal image restoration tasks.