ADAPTIVE EDGE SELECTION FOR BLIND IMAGE DEBLURRING

Abstract

Blind image deblurring remains to be a challenging problem in the field of image restoration, largely due to the difficulty in estimating accurate motion blur kernels, especially when the blur is large and the observed image lack useful information for kernel estimation.Based on the observation that a good intermediate estimate of latent image for estimating motion blur kernel is not necessarily the one closest to latent image, edge selection has proven itself a very powerful technique for achieving state-of-the-art performance in blind image deblurring. However, there is few clear guidance on how to perform edge selection in order to achieve better accuracy for kernel estimation. In this thesis, we will explore the use of machine leaning algorithms in image deblurring tasks. Two new deblurring algorithms that use machine leaning algorithms to adaptively select edges for kernel estimation will be introduced.