ENHANCING UNDEREXPOSED IMAGES USING U-NET AND DEEP LEARNING TECHNIQUE

Vinod Kumar Dhiman*

This research addresses the challenge of enhancing underexposed images in the context of computer vision and photography. Under- exposed images, characterized by low levels of captured photons, pose significant limitations for various applications, including autonomous vehicles, medical imaging, and photography. I propose a novel approach that leverages the U-Net architecture, a convolutional neural network known for its effectiveness in image-processing tasks. Our model is designed to recover underexposed images while preserving sharpness, color accuracy, and cleanliness. I also explore integrating generative adversarial networks (GANs) and attention mechanisms to improve image quality further. Through extensive experimentation and evaluation 1, I demonstrate the potential of our method in significantly enhancing underexposed images, making them suitable for critical applications such as object detection, medical diagnosis, and photography. Future research directions include addressing model generalization across sensors and optimizing real-time processing capabilities.