Constrained Nonlocal Total Variational Deblurring Models And Fast Algorithms Based On ADMM

This paper focuses on the constrained NLTV deblurring models using ADMM.NLTV is originated from Total Variation Model and performs better in image deblurring compared with TV models.As we all know,in image processing,the numerical value of pixel point usually stays in a fixed range.Therefore,in algorithm,we need to project the minimizer back into the range,which causes that the value after projection will not be the minimizer again and the accuracy of the algorithm will be affected.In this paper we established the constrained model.By adding constraints and solving corresponding subproblems,we used ADMM on constrained NLTV model and resolved the contradiction between projection and minimization.The numerical result shows that we gain better performance in contrast to some other algorithms.