Restoring Image Clarity: Dual-GAN Framework for High-Quality Image Generation from Low-Resolution Blurred Inputs

Neeraja s, George Mathew

Abstract

A key task in computer vision is image restoration, which aims to recover high-quality
images from blurry or degraded inputs. In this research, we provide a novel method for
restoring high-quality images by combining the advantages of DeblurGAN and
ESRGAN (Enhanced Super-Resolution GAN). While ESRGAN focuses on superresolution and fine detail recovery, DeblurGAN focuses on eliminating blurring
artefacts and increasing image details. We take advantage of the complementary
strengths of these two architectures and produce a synergistic outcome by combining
them in a multi-stage refining process. The suggested method starts by using
DeblurGAN to deblur the input photos and improve their clarity and sharpness. Enhance
super-resolution GAN is then used to further enhance the deblurred images with a focus
on high-frequency detail recovery and super-resolution. The input photos can be fully
restored using this multi-stage refinement procedure, which improves visual quality and
resolution. Numerous tests show that this method performs better in eliminating
blurring artefacts and retrieving fine details. Deblur-GAN and ESRGAN are combined
to improve visual quality, resolution, and detail recovery through a multi-stage
refinement process. The method effectively addresses the picture restoration issues,
showing potential for applications in digital photography, surveillance, and medical
imaging.