LLCNN-based Image Enhancement with Squeeze and Excitation Network for Improved Visual Quality

Image enhancement, particularly in low-light conditions, has long been a research focus to improve visual quality. Low-light settings often cause visibility issues and detail loss, which recent deep-learning techniques have effectively addressed. This study explores the combination of a low-light convolutional neural network (LLCNN), a convolutional neural network designed explicitly for low-light image enhancement, with a Squeeze and Excitation Network (SEN), which recalibrate features to emphasise critical details and suppress noise. Integrating SEN into LLCNN enhances image details while reducing noise and artefacts typically present in low-light conditions. Experimental results show that this combined approach outperforms existing methods in objective metrics, such as peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM), as well as visual quality.