Gan model for image distortion recovery system

Abstract

Image distortion is a common problem that affects the quality and appearance of digital images. Regular image restoration techniques and algorithms to get back the expected stage of pixels in not that much accurate & effective as it works on the restoration of the already damaged pixel which cannot be considered as an effective method. A research on the use of Pix2Pix and U-NET for the restoration of the image has been presented in this paper. In this research paper we developed and improved image restoration method to where the Pix2Pix GAN model can learn a mapping from distorted images to their corresponding undistorted versions, resulting in high- quality image restoration. The proposed method has the potential to improve the visual quality and content of degraded images in many applications, including medical imaging and remote sensing. This model works on the concept of regenerating a whole new image to remove distortion as it is an improved method. At last, the performance of the Pix2Pix and GAN is evaluated to validate the model and this shows the improved effectiveness of the system and its capabilities. In future these models can be trained to the extent where they can generate images on their own much faster than the traditional image restoration methods to give de blurred images.