[Paper Review] A Total Fractional-Order Variation Model for Image Restoration with Non-homogeneous Boundary Conditions and its Numerical Solution

Jianping Zhang, Ke Chen|arXiv (Cornell University)|Sep 6, 2015

Fractional Differential Equations Solutions54 references3 citations

TL;DR

This paper proposes a novel fractional-order total variation model for image restoration that uses non-homogeneous boundary conditions to overcome limitations of traditional total variation and high-order models. By rigorously analyzing the theoretical properties and developing efficient numerical algorithms, the method achieves superior restoration quality and efficiency, outperforming established models like mean curvature and total generalized variation in preserving smoothness and reducing staircasing effects.

ABSTRACT

To overcome the weakness of a total variation based model for image restoration, various high order (typically second order) regularization models have been proposed and studied recently. In this paper we analyze and test a fractional-order derivative based total $α$-order variation model, which can outperform the currently popular high order regularization models. There exist several previous works using total $α$-order variations for image restoration; however first no analysis is done yet and second all tested formulations, differing from each other, utilize the zero Dirichlet boundary conditions which are not realistic (while non-zero boundary conditions violate definitions of fractional-order derivatives). This paper first reviews some results of fractional-order derivatives and then analyzes the theoretical properties of the proposed total $α$-order variational model rigorously. It then develops four algorithms for solving the variational problem, one based on the variational Split-Bregman idea and three based on direct solution of the discretise-optimization problem. Numerical experiments show that, in terms of restoration quality and solution efficiency, the proposed model can produce highly competitive results, for smooth images, to two established high order models: the mean curvature and the total generalized variation.

Motivation & Objective

• To address the blocky (staircase) effect and contrast loss in total variation-based image restoration models.
• To develop a theoretically sound fractional-order total α-order variation model that generalizes classical and high-order regularization.
• To resolve the inconsistency in prior works by properly handling non-zero (non-homogeneous) boundary conditions for fractional derivatives.
• To design efficient numerical algorithms for solving the resulting variational problem with robust convergence.
• To demonstrate competitive performance against state-of-the-art high-order models in image denoising and restoration quality.

Proposed method

• Proposes a total α-order variation model based on Riemann-Liouville fractional derivatives for image regularization, with α ∈ (1,2) to balance smoothness and edge preservation.
• Introduces a variational formulation that couples data fidelity and fractional-order regularization, minimizing the energy functional via constrained optimization.
• Applies the Split-Bregman method to decouple the optimization problem into subproblems, enabling efficient numerical solution.
• Develops three alternative algorithms based on direct solution of the discretized optimization problem, enhancing robustness and convergence.
• Incorporates non-homogeneous boundary conditions by deriving necessary conditions using the Caputo fractional divergence, ensuring consistency with physical and mathematical definitions.
• Uses Gateaux derivative and integration by parts to derive the Euler-Lagrange equation, with boundary terms handled via fractional-order trace conditions.

Experimental results

Research questions

• RQ1 Can a fractional-order total variation model with non-homogeneous boundary conditions outperform classical and high-order regularization models in image restoration?
• RQ2 How can fractional-order derivatives be consistently applied in image variational models when non-zero boundary conditions are required?
• RQ3 What are the theoretical properties of the proposed total α-order variation model, particularly in terms of existence and uniqueness of solutions?
• RQ4 How do different numerical algorithms compare in terms of convergence speed, stability, and restoration quality?
• RQ5 To what extent does the proposed model reduce the staircasing effect while preserving image contrast and fine structures?

Key findings

• The proposed fractional-order model achieves restoration quality comparable to or better than the mean curvature and total generalized variation models, especially for smooth images.
• Numerical experiments show that the PDE-Split-Bregman (PDE-SB) algorithm outperforms other solvers in both restoration accuracy and computational efficiency.
• The model effectively mitigates the staircasing effect common in total variation while preserving image contrast and fine textures.
• Theoretical analysis confirms the well-posedness of the variational problem under non-homogeneous boundary conditions, resolving inconsistencies in prior works.
• Boundary terms in the fractional derivative formulation are properly handled via Caputo-type divergence, ensuring mathematical consistency.
• The method demonstrates robustness across various noise levels and image types, with stable convergence across multiple test cases.