AN EVALUATION STUDY OF THE NAFNET MODEL FOR IMAGE DEBLURRING AND DENOISING

Abstract

Image restoration is a fundamental problem in image processing and computer vision, with applications in security surveillance, medical imaging, and intelligent vision systems. Recent deep learning–based approaches have achieved significant improvements. Among them, NAFNet is a recently proposed architecture that removes nonlinear activation functions to achieve a simple and efficient design. This paper evaluates the performance of NAFNet and compares it with several state-of-the-art image restoration models, including DeepDeblur, SRNNet, and MPRNet, experimental results are obtained on benchmark datasets using PSNR and SSIM as evaluation metrics. The results indicate that NAFNet achieves superior restoration quality, preserves fine image details, converges faster during training, and demonstrates competitive overall performance compared to existing methods.