Inertial Nonconvex Alternating Minimizations for the Image Deblurring.

Area of Science:

• Image Processing
• Numerical Analysis
• Optimization

Background:

• Total Variation (TV) regularization is crucial for image deblurring.
• Alternating Direction Method of Multipliers (ADMM) is a popular convex optimization solver.
• Nonconvex formulations are gaining traction for improved numerical performance in image reconstruction.

Purpose of the Study:

• To propose a novel nonconvex alternating direction method of multipliers with inertial technique (IADMM).
• To extend classical TV regularization models using nonconvex optimization.
• To analyze the convergence properties of the proposed IADMM algorithm.

Main Methods:

• Development of the Inertial Alternating Direction Method of Multipliers (IADMM).
• Application of the Kurdyka-Łojasiewicz property to prove algorithm convergence.
• Numerical simulations for classical TV image reconstruction tasks.

Main Results:

• The proposed IADMM algorithm shows efficient performance in TV image reconstruction.
• Convergence of the IADMM is proven under specific parameter assumptions.
• Numerical comparisons highlight the behavior of IADMM against established ADMM methods.

Conclusions:

• The novel IADMM offers an efficient approach for nonconvex TV image reconstruction.
• The study validates the convergence and practical utility of the IADMM algorithm.
• This work contributes to advancing optimization techniques in image processing.