Encoder-Only Image Registration

2025-08-30

AI Summary

To address high computational complexity and difficulty in modeling large deformations in deformable image registration, this paper proposes an efficient diffeomorphic registration framework. Methodologically, it introduces a lightweight encoder-decoder architecture with feature separation, employing only three convolutional layers for feature extraction and a Laplacian feature pyramid for progressive deformation modeling; a multi-scale optical flow estimator synthesizes diffeomorphic deformation fields across three resolution levels. Crucially, the work is the first to reveal that convolutional networks inherently perform dual functions in registration: local intensity linearization and global contrast harmonization. Evaluated on five cross-modality and cross-anatomical datasets, the method achieves state-of-the-art trade-offs between accuracy and efficiency, as well as accuracy and deformation smoothness, demonstrating significant improvements in both computational speed and diffeomorphic regularity while maintaining high registration accuracy.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning with Manifolds
• Search and Optimization: Non-convex Optimization

Application Category

• 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
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Learning-based techniques have significantly improved the accuracy and speed of deformable image registration. However, challenges such as reducing computational complexity and handling large deformations persist. To address these challenges, we analyze how convolutional neural networks (ConvNets) influence registration performance using the Horn-Schunck optical flow equation. Supported by prior studies and our empirical experiments, we observe that ConvNets play two key roles in registration: linearizing local intensities and harmonizing global contrast variations. Based on these insights, we propose the Encoder-Only Image Registration (EOIR) framework, designed to achieve a better accuracy-efficiency trade-off. EOIR separates feature learning from flow estimation, employing only a 3-layer ConvNet for feature extraction and a set of 3-layer flow estimators to construct a Laplacian feature pyramid, progressively composing diffeomorphic deformations under a large-deformation model. Results on five datasets across different modalities and anatomical regions demonstrate EOIR's effectiveness, achieving superior accuracy-efficiency and accuracy-smoothness trade-offs. With comparable accuracy, EOIR provides better efficiency and smoothness, and vice versa. The source code of EOIR will be publicly available on https://github.com/XiangChen1994/EOIR.

Problem

• Reducing computational complexity in image registration
• Handling large deformations in deformable registration
• Achieving better accuracy-efficiency trade-off in registration

Innovation

• Encoder-Only framework separates feature learning
• Uses Laplacian pyramid for progressive deformation composition
• 3-layer ConvNet achieves accuracy-efficiency trade-off