Blind Image Restoration Based on Extreme Channel Prior and Gradient Cepstrum
YU Lun
XING Xiaoxiao
Abstract：Aiming at the problems of ringing artifacts in the images recovered by the extreme channel prior deblurring method,a single image blind deblurring method based on the extreme channel prior and gradient cepstrum is proposed.Firstly,the l0 norm constraint is imposed on the extreme channel prior and introduced into the maximum a posteriori probability framework to construct the fuzzy kernel estimation model.Then,the fuzzy kernel is estimated by multi-scale alternating iterative estimation,and the semi-quadratic splitting method is used to effectively solve the non-convex problem of the model in the iterative process.In order to inhibit the excessive iteration of each scale,the kernel similarity is used to evaluate the subtle changes of the fuzzy kernel in the iter-ative process,so that the final iterative fuzzy kernel is more accurate.Finally,the image deblurring is realized by non-blind deconvo-lution.Experimental results show that the proposed method achieves good results on both synthetic and real datasets,which can sup-press artifacts and recover more image details.
Keywords：extreme channel priorgradient cepstrumfuzzy kernelmulti-scalehalf quadratic splitting
Publication Date：2024-10-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 31-36 )