Improved ADMM based TV minimized Image deblurring without boundary artifacts

Edge preserving (e.g. total variation minimized) regularized image deblurring methods are actively researched with many practical applications. Formally this type of deblurring is equivalent to a convex non-smooth optimization problem. In this paper we describe an ADMM optimization based effective a...

Full description

Saved in:
Bibliographic Details
Published in2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC) pp. 002000 - 002005
Main Authors Hadhazi, Daniel, Horvath, Aron, Horvath, Gabor
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2016
Subjects
ADMM
Concrete
Convergence
Edge preserving regularization
Image restoration
Imaged deblurring
Linear programming
Optimization methods
Total variation minimization
Online AccessGet full text

Cover

More Information
Summary:Edge preserving (e.g. total variation minimized) regularized image deblurring methods are actively researched with many practical applications. Formally this type of deblurring is equivalent to a convex non-smooth optimization problem. In this paper we describe an ADMM optimization based effective algorithm, which can be considered as an improved version of a previously published method. Due to the introduced modifications different loss functions can be easily used, positivity constraint is applied and the speed of the convergence of deblurring is also increased. The quality of the deblurring in the cases of using different loss functions is compared qualitatively and quantitatively and the accelerating rate of the convergence is also examined. For these measurements benchmark images were used. Based on our experiments we suggest to consider Huber function as loss function instead of the commonly used quadratic functions.
DOI:10.1109/SMC.2016.7844534