l₀TV: A Sparse Optimization Method for Impulse Noise Image Restoration.

Area of Science:

• Image Processing
• Computer Vision
• Optimization

Background:

• Impulse noise is a common artifact in image acquisition and transmission.
• Existing Total Variation (TV) based methods, like Adaptive Outlier Pursuit (AOP), offer sub-optimal performance for impulse noise removal.
• Effective impulse noise reduction is crucial for image restoration tasks.

Purpose of the Study:

• To propose a new sparse optimization method, l0TV-PADMM, for image restoration with impulse noise.
• To address the challenges posed by non-convex and non-smooth optimization problems in TV-based restoration.
• To improve upon the performance of state-of-the-art methods in impulse noise removal.

Main Methods:

• Developed l0TV-PADMM, a novel method utilizing Total Variation (TV) with l0-norm data fidelity.
• Reformulated the non-convex optimization problem into a biconvex Mathematical Program with Equilibrium Constraints (MPEC).
• Employed a proximal Alternating Direction Method of Multipliers (PADMM) to solve the reformulated problem.

Main Results:

• l0TV-PADMM effectively solves the l0-norm optimization problem for image restoration.
• The method demonstrates convergence under mild conditions.
• Experimental results show l0TV-PADMM outperforms existing state-of-the-art methods in image denoising and deblurring with impulse noise.

Conclusions:

• l0TV-PADMM is a robust and effective method for impulse noise removal in image restoration.
• The proposed approach offers significant improvements over current techniques.
• This work advances the field of regularization-based image processing for noisy image restoration.