Image quality enhancement remains a critical challenge in image processing techniques, especially for images captured under suboptimal lighting conditions that can generate low contrast, color distortion, and noise. These degradations not only impact visual aesthetics but also hinder the performance of downstream image processing tasks. The primary objective of image quality enhancement is to improve the visual quality of such images to benefit subsequent processing. Despite extensive research, achieving high-quality enhanced images remains challenging. Traditional image quality enhancement techniques often address only overexposure or underexposure, potentially failing when both issues are present. Deep learning has recently been increasingly adopted in image processing, demonstrating significant potential for enhancing image quality with underexposure, overexposure, or a combination of both. In this paper, we first review key traditional and machine learning-based image quality enhancement techniques developed in recent years. Next, we contribute by creating a new dataset to facilitate learning. We then propose an improved denoising step for input images and integrate two Residual Blocks into the “Color Shift Estimation and Correction” network architecture to enhance feature extraction. Furthermore, we introduce a novel loss function, \(MSE_{LOSS}\) , aimed at ensuring both pixel-level accuracy and perceptual realism in the enhanced images, which leads to improved visual quality. Finally, we employ a Color Shift Estimation and Correction method to train our model using both public and our newly constructed dataset. Experimental results on our new dataset demonstrate the effectiveness of our proposed approach in generating high-quality enhanced images with improved color fidelity and well-preserved details.