Denoising of Aged Album and Mural Photographs Using GAN Techniques

This particular study introduces a GAN (Generative Adversarial Network) based methodology to denoise historic photographs. Different forms of Speckle., some form of Gaussian impurity., Salt and Pepper noise., frequently damages the historical photographic clarity. These kinds of noise might make it harder to see crucial aspects of an image and impede additional image processing. We provide a denoising approach based on a GAN to solve this problem. These types of GAN's are used to generate denoised image from a noisy input after being trained to learn the underlying clean image distribution. The testing findings show that the suggested strategy may successfully eliminate various types of noise from old photographs and murals and enhance the quality visually. The outputs demonstrate that the suggested methodology outperforms the other denoising techniques based on the visual quality and denoising efficacy. We have used a total of 550 samples on both the GAN models. For Pix2Pix GAN., we managed to achieve PSNR of 31.295., MSE of 0.571 and SSIM of 0.613 at 500 epochs. GFP GAN., a pre-trained model., was able to obtain PSNR of 33.580., MSE of 0.057 and SSIM of 0.892. This work offers a viable method for denoising old photos and could make an impact on many different image processing applications.