RBE-Flow: Recurrent Bayesian Estimation on Feature Manifolds for Cross-Modal Registration

2026-06-29

AI Summary

This work addresses the challenges of matching ambiguity and error accumulation in cross-modal image registration caused by nonlinear radiometric discrepancies and geometric distortions. The authors propose a closed-loop recursive Bayesian optical flow estimation method grounded in feature manifolds. By jointly optimizing feature metrics and probabilistic state updates, the approach integrates recursive manifold optimization with an uncertainty-adaptive mechanism that dynamically adjusts optimization damping via posterior covariance. A geometry-aware loss function is further introduced to prevent variance collapse. Leveraging deterministic sigma-point projection and a hybrid supervised training strategy, the method achieves state-of-the-art performance on benchmark datasets—including OSdataset, WHU-OPT-SAR, and RoadScene—with particularly notable gains in sub-pixel accuracy.

Abstract

Cross-modal image registration is essential for multi-sensor perception but remains fundamentally challenging due to severe non-linear radiometric discrepancies and geometric distortions. Existing deterministic matching methods lack uncertainty awareness, struggling to navigate the resulting highly non-convex optimization landscape and frequently accumulating errors in ambiguous regions. In this paper, we propose RBE-Flow, a novel framework that reformulates dense cross-modal flow estimation as a closed-loop recurrent Bayesian estimation problem on learned feature manifolds. Diverging from standard feed-forward regression, RBE-Flow establishes a robust self-correcting mechanism by deeply coupling feature-metric non-linear optimization with probabilistic state updates. Specifically, a Recurrent Manifold Optimization (RMO) block iteratively generates flow observations and their associated uncertainties, which are then optimally assimilated into the prior state via an Uncertainty-Adaptive Probabilistic Update (UAPU) using deterministic sigma-point projection. Crucially, the resulting calibrated posterior covariance is fed back to adaptively regularize the damping of subsequent optimization steps, allowing the system to modulate its convergence based on predictive confidence. To ensure stable probabilistic training, we introduce a hybrid supervision scheme featuring a geometry-aware rectified NLL loss that structurally prevents variance collapse. Extensive experiments on challenging OSdataset, WHU-OPT-SAR, and RoadScene benchmarks demonstrate that RBE-Flow consistently achieves state-of-the-art performance, outperforming existing methods by a significant margin, particularly under strict sub-pixel criteria. Project page: https://github.com/NEU-Liuxuecong/RBE-Flow

Problem

• cross-modal registration
• non-linear radiometric discrepancies
• geometric distortions
• uncertainty awareness
• non-convex optimization

Innovation

• Recurrent Bayesian Estimation
• Feature Manifolds
• Uncertainty-Aware Registration
• Cross-Modal Flow
• Probabilistic Optimization