EDFFDNet: Towards Accurate and Efficient Unsupervised Multi-Grid Image Registration

2025-09-09

AI Summary

To address degraded image registration accuracy caused by depth variation in real-world scenarios, this paper proposes an efficient unsupervised multi-grid registration method. The approach introduces (1) exponential-decay free-form deformation (ED-FFD) to model non-rigid deformations; (2) an adaptive sparse motion aggregator (ASMA) for robust motion field estimation; and (3) a global-to-local progressive correlation refinement mechanism. The method supports end-to-end unsupervised training. Compared to state-of-the-art baselines, it reduces parameter count, memory footprint, and runtime by 70.5%, 32.6%, and 33.7%, respectively, while improving PSNR by 0.5 dB; incorporating local optimization further boosts PSNR by 1.06 dB. It demonstrates significantly superior generalization over existing homography-, TPS-, and multi-grid-based methods.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Multimodal Learning
• Multiagent Systems: Adversarial Agents

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Previous deep image registration methods that employ single homography, multi-grid homography, or thin-plate spline often struggle with real scenes containing depth disparities due to their inherent limitations. To address this, we propose an Exponential-Decay Free-Form Deformation Network (EDFFDNet), which employs free-form deformation with an exponential-decay basis function. This design achieves higher efficiency and performs well in scenes with depth disparities, benefiting from its inherent locality. We also introduce an Adaptive Sparse Motion Aggregator (ASMA), which replaces the MLP motion aggregator used in previous methods. By transforming dense interactions into sparse ones, ASMA reduces parameters and improves accuracy. Additionally, we propose a progressive correlation refinement strategy that leverages global-local correlation patterns for coarse-to-fine motion estimation, further enhancing efficiency and accuracy. Experiments demonstrate that EDFFDNet reduces parameters, memory, and total runtime by 70.5%, 32.6%, and 33.7%, respectively, while achieving a 0.5 dB PSNR gain over the state-of-the-art method. With an additional local refinement stage,EDFFDNet-2 further improves PSNR by 1.06 dB while maintaining lower computational costs. Our method also demonstrates strong generalization ability across datasets, outperforming previous deep learning methods.

Problem

• Addresses multi-grid image registration with depth disparities
• Reduces computational parameters and memory usage
• Improves registration accuracy and efficiency

Innovation

• Exponential-decay free-form deformation for depth disparities
• Adaptive sparse motion aggregator reduces parameters
• Progressive correlation refinement for coarse-to-fine estimation