Non-Blind and Blind Deconvolution Methodologies in Restoration of Motion-Blurred images

Image blurring due to factors like light, motion, and humidity presents a significant challenge in image processing, particularly in practical applications like satellite imaging, medical imaging, and sports photography, where the Point Spread Function (PSF) is often unknown. Several non-blind and advanced deep learning techniques have been suggested to tackle this problem. This study explores various blind and non-blind deconvolution methods for deblurring motion-blurred images, with a particular focus on scenarios where the PSF remains unknown. Blind deconvolution offers a powerful approach in such cases. The proposed autoencoder-CNN framework excels at deblurring motion blur images. The experimental evaluation, based on an extensively collected dataset, demonstrates that blind deconvolution techniques outperform non-blind methods for all types of degradation, especially for heavily degraded images.