Optical flow with fractional order regularization: variational model and solution method

TL;DR

This paper introduces a novel optical flow model using fractional order regularization and a split Bregman solution method, demonstrating improved performance and analyzing the impact of fractional order on estimation accuracy.

Contribution

The paper proposes a new variational optical flow model with fractional differential regularization and develops an efficient split Bregman algorithm for its solution.

Findings

The new model performs favorably compared to existing methods.

The effectiveness of fractional order regularization depends on image complexity.

Optimal fractional order varies with image geometry and texture.

Abstract

An optical flow variational model is proposed for a sequence of images defined on a domain in $\mathbb{R}^2$. We introduce a regularization term given by the $L^1$ norm of a fractional differential operator. To solve the minimization problem we apply the split Bregman method. Extensive experimental results, with performance evaluation, are presented to demonstrate the effectiveness of the new model and method and to show that our algorithm performs favorably in comparison to another existing method. We also discuss the influence of the order $\alpha$ of the fractional operator in the estimation of the optical flow, for $0 \leq \alpha \leq 2$. We observe that the values of $\alpha$ for which the method performs better depends on the geometry and texture complexity of the image. Some extensions of our algorithm are also discussed.