Blind and Non-Blind Deconvolution-Based Image Deblurring Techniques for Blurred and Noisy Image

Abstract: Image deblurring is a common issue in low-level computer vision aiming to restore a clear image from a blurred input image. Deep learning innovations have significantly advanced the solution to this issue, and numerous deblurring networks have been presented to recover high-quality images. This study aims to investigate the impact of Blind deconvolution and Non-Blind Deconvolution (Weiner Filter, Regularized Filter, and lucky Richardson) deblurring techniques and blind deconvolution to retrieve the original image from the blurring and the noisy images. Point Spread Function (PSF) is required to perform the deconvolution process. MATLAB program is utilized in this study as a suitable tool for image processing. Peak to Signal Ratio (PSNR) and structural index similarity (SSIM) are the major parameters used to examine image quality. The results showed that the Regularized Filter was an effective technique to deblur the blurry image, and it achieved the largest PSNR and best SSIM with the prior information about the PSF for different degrees of blurring angle. These four deblurring techniques were unsuccessful in restoring the original image from the image with Gaussian noise.