An iterative shrinkage approach to total-variation image restoration

Area of Science:

• Computer Vision
• Image Processing
• Applied Mathematics

Background:

• Digital image restoration is crucial for many applications but challenging with ill-conditioned operators and noise.
• A priori information (priors) is essential for stable and robust image recovery, with Total Variation (TV) being a standard for piecewise smooth images.
• Existing TV-based restoration algorithms are numerous, motivating the search for more efficient and effective methods.

Purpose of the Study:

• To propose a novel, efficient algorithm for Total Variation (TV)-based image restoration.
• To develop a method that works directly with the TV functional, unifying isotropic and anisotropic definitions.
• To demonstrate superior performance in image deblurring compared to existing techniques.

Main Methods:

• A new approach using iterative shrinkage (iterated thresholding) for TV-based image restoration.
• The method recursively applies linear filtering and soft thresholding.
• It is classified as a first-order algorithm, suitable for large-scale image processing.

Main Results:

• The proposed iterative shrinkage method achieves high-quality image restoration.
• It directly utilizes the TV functional, avoiding smoothed approximations.
• The method offers a unified solution for both isotropic and anisotropic TV definitions.
• Demonstrated superior deblurring results compared to sparse-wavelet deconvolution in standard examples.

Conclusions:

• The iterative shrinkage method provides an efficient and effective solution for TV-based image restoration.
• It offers a unified treatment of isotropic and anisotropic TV functionals.
• The technique yields superior restoration quality, particularly in challenging deblurring scenarios.