Blur Kernel

Blur kernel estimation and removal are central to image restoration, aiming to recover sharp images from blurry ones, particularly when the blurring process is unknown (blind deblurring). Current research heavily utilizes deep learning, employing diverse architectures like diffusion models, generative adversarial networks, and recurrent convolutional networks, often incorporating alternating optimization strategies to jointly estimate the blur kernel and the sharp image. These advancements significantly improve image quality and efficiency in applications ranging from remote sensing to computational photography, addressing limitations of traditional model-based methods.