Enhanced SURF-Based Image Registration for Optoelectronic Sensors Using Circular Descriptors and Bidirectional Geometric Constraints
Journal of Nanoelectronics and Optoelectronics · 01.02.2025 · 20 (2) · 164-172
DOI: 10.1166/jno.2025.3735 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Nanoelectronics and Optoelectronics, erschienen am 01.02.2025. Verfasst von Xin Ye und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Image registration is a critical task in optoelectronic imaging systems, where accurate alignment of overlapping images is essential for applications such as remote sensing, medical imaging, and industrial inspection. While feature-based methods like Speeded-Up Robust Features (SURF) offer efficient registration, challenges remain in balancing computational speed and matching accuracy. This paper presents an improved SURF-based image registration method optimized for optoelectronic image sensors. The proposed approach introduces a circular neighborhood for descriptor construction, replacing the conventional square region to enhance rotational invariance while reducing descriptor dimensionality from 64 to 20, significantly lowering computational complexity. Additionally, an adaptive bidirectional K-Nearest Neighbors (KNN) matching strategy is employed to improve initial matching accuracy, followed by a refined Random Sample Consensus(RANSAC) algorithm incorporating bidirectional geometric constraints to eliminate mismatches. Experimental results demonstrate that the proposed method achieves a matching accuracy rate (MAR) of 87.37%, outperforming traditional SURF (83.15%) and ORB (71.23%) while maintaining near-SIFT-level precision (89.56%) at substantially reduced computational cost. Execution time comparisons reveal that the enhanced algorithm operates 27% faster than classical SURF and over 75% faster than SIFT, making it highly suitable for realtime optoelectronic imaging applications. The improvements in both efficiency and robustness highlight the method’s potential for high-performance image registration in nanoelectronics and optoelectronics, particularly in scenarios demanding rapid processing without sacrificing accuracy.
Zitieren
APA
Ye, X., & Yang, F. (2025). Enhanced SURF-Based Image Registration for Optoelectronic Sensors Using Circular Descriptors and Bidirectional Geometric Constraints. Journal of Nanoelectronics and Optoelectronics, 20(2), 164-172. https://doi.org/10.1166/jno.2025.3735
Vancouver
Ye X, Yang F. Enhanced SURF-Based Image Registration for Optoelectronic Sensors Using Circular Descriptors and Bidirectional Geometric Constraints. Journal of Nanoelectronics and Optoelectronics. 2025;20(2):164-172. doi:10.1166/jno.2025.3735
BibTeX
@article{Ye2025,
author = {Ye, Xin and Yang, Fan},
title = {Enhanced SURF-Based Image Registration for Optoelectronic Sensors Using Circular Descriptors and Bidirectional Geometric Constraints},
journal = {Journal of Nanoelectronics and Optoelectronics},
year = {2025},
volume = {20},
number = {2},
pages = {164--172},
doi = {10.1166/jno.2025.3735}
}
EUVIMED