Overcoming Complex Transformations in Image Registration: Methods for Enhanced Accuracy

Image registration is a critical process across diverse disciplines such as remote sensing, medical imaging, and cultural heritage conservation. Robust registration methods underpin accurate image comparison, fusion, and change detection—tasks essential for diagnostics, monitoring, and preservation. However, complex transformations, ranging from large-scale deformations and nonlinear spatial changes to varying intensity patterns, pose enduring challenges to practitioners. Traditional registration algorithms often struggle in these scenarios, especially under real-world conditions marked by noise, artifacts, or anatomical variability. Recent advances, including phase congruency models, hybrid optimization frameworks, and adaptive similarity measures, have demonstrated promising improvements. Yet, achieving reliable and accurate alignment under complex transformations remains a pressing open problem.GoalsThis Research Topic aims to drive innovation in image registration by soliciting research that addresses the core challenges of robust alignment under complex transformations. The Topic seeks contributions that push the boundaries of algorithmic robustness and accuracy, whether through advanced feature extraction, improved noise suppression, or novel optimization strategies. Key goals include developing methods that generalize across imaging modalities or datasets, as well as providing benchmarking studies and validation frameworks for real-world applications. Of particular interest are solutions that enhance performance in settings prone to large deformations, local and global intensity variations, or heterogeneous spatial structures.Scope and Information for AuthorsInterdisciplinary submissions from computer vision, medical imaging, remote sensing, digital heritage, and related fields are welcome. Relevant themes and topics include but are not limited to:• Algorithmic advances for complex transformations: Novel methods, including deep learning, variational approaches, or hybrid techniques, tailored for nonlinear, nonrigid, or multi-modal registration problems.• Feature extraction and matching under challenging conditions: Strategies that improve robustness to noise, artifacts, or low signal-to-noise ratios (SNR).• Similarity measures and validation metrics: New objective functions, benchmarking protocols, or statistical frameworks for evaluating registration accuracy where ground truth is uncertain.• Application-driven studies: Demonstrations of registration pipelines in domains such as longitudinal medical imaging, remote sensing of environmental changes, or the analysis of historic art and artifacts.• Open datasets and reproducible benchmarking: Contributions that introduce or curate datasets and share resources for standardized evaluation.Through this Research Topic, the intention is to bring together foundational and applied research that advances the reliability and effectiveness of image registration techniques—ultimately supporting innovation across scientific, engineering, and cultural domains. Topic editors: Jaime Zabalza, Indranil Misra, Yipeng Qin Source: https://www.frontiersin.org/research-topics/72825/overcoming-complex-transformations-in-image-registration-methods-for-enhanced-accuracy