K-GMRF: Kinetic Gauss-Markov Random Field for First-Principles Covariance Tracking on Lie Groups

TL;DR

K-GMRF introduces a second-order, Lie group-based covariance tracking method that outperforms existing estimators by reducing phase lag and achieving zero steady-state error, applicable across vision tasks.

Contribution

The paper presents K-GMRF, a novel second-order covariance tracking framework on Lie groups derived from Euler-Poincaré equations, improving accuracy and stability over first-order methods.

Findings

Reduces angular error by 30x on synthetic ellipses

Decreases geodesic error from 29.4° to 9.9° under dropout

Improves loU from 0.55 to 0.74 on BlurCar2 sequences

Abstract

Tracking non-stationary covariance matrices is fundamental to vision yet hindered by existing estimators that either neglect manifold constraints or rely on first-order updates, incurring inevitable phase lag during rapid evolution. We propose K-GMRF, an online, training-free framework for covariance tracking that reformulates the problem as forced rigid-body motion on Lie groups. Derived from the Euler-Poincar\'e equations, our method interprets observations as torques driving a latent angular velocity, propagated via a structure-preserving symplectic integrator. We theoretically prove that this second-order dynamics achieves zero steady-state error under constant rotation, strictly superior to the proportional lag of first-order baselines. Validation across three domains demonstrates robust tracking fidelity: (i) on synthetic ellipses, K-GMRF reduces angular error by 30x compared to Riemannian EMA while maintaining stability at high speeds; (ii) on SO(3) stabilization with 20% dropout, it decreases geodesic error from 29.4{\deg} to 9.9{\deg}; and (iii) on OTB motion-blur sequences, it improves loU from 0.55 to 0.74 on BlurCar2 with a 96% success rate. As a fully differentiable symplectic module, K-GMRF provides a plug-and-play geometric prior for data-constrained scenarios and an interpretable layer within modern deep architectures.