Fuzzy Image Restoration and Detail Enhancement using Improved Generative Adversarial Networks

Recently, images are frequently damaged due to data loss, and blurring obtained by noise occurrences and sampling. Image restoration and detail enhancement purpose is to retrieve a sharp image from the source of the blurred image, which has been a basic problem and challenging in image processing and computer vision. Hence, in this research Improved Generative Adversarial Networks (IGANs) is proposed to generate high-quality realistic images with accurate details for better visual image representation systems. Initially, Levins dataset is used to collect the data of the image. Then, image decomposition separates an image into different allowing targeted processing to improve clarity and enhance fine details effectively. The Fuzzy image restoration and detail enhancement separating an image to process each part optimally, improving overall quality and fine details. Finally, the proposed IGANs is utilized to accuracy of image representation systems for better visual results. According the results it illustrates when compared to existing models such as Dual-Branch Sparse Priors GAN (DBSGAN), Deconvolutional De-Blurring Algorithm (DDA) and Distortion-Corrected Reconstruction Network (DCReconNet) the recommended IGANs strategy outperformed the others in terms of SSIM of 0.995, AND PSNR of 15.12 dB respectively.