Enhancing Image Clarity using Deep Dehazing Architecture

Image dehazing is an imperative task in computer vision and research and development, it has various applications in many different sectors, including autonomous driving and surveillance. In proposed work, we present a different strategy for dehazing images using Generative Adversarial Networks (GAN) and U-Net. Our proposed model consists of a U-Net-based network and a discriminator network that distinguishes the generated and actual image. The U-Net-based network receives an input image which is a hazy and generates output dehazed image, while the discriminator network is trained to differentiate between true and fake images. By using a Generative Adversarial Networks Sustainability, we can confirm that the output images are more realistic and visually pleasing, while the U-Net architecture helps preserve important details and structures in the image with research precision.