EUVIMEDEUVIMED entdecken
Crossref

Multispectral infrared–visible image stitching based on transformer architecture and feature complementarity

Xue Lu, Chunhui Niu, Ming Hao

Applied Optics · 20.07.2026 · 65 (21) · 7028

DOI: 10.1364/ao.600053 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Applied Optics, online erschienen am 13.07.2026. Verfasst von Xue Lu und 2 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To address the degradation of stitching performance caused by environmental interference and information loss in image stitching tasks, we propose a transformer-based infrared–visible image stitching method with a feature complementarity strategy. The proposed method consists of a progressive regression network for image pre-alignment and an attention-guided network for image reconstruction. In the pre-alignment stage, a hybrid attention mechanism is introduced to enhance cross-modal feature representation, while disparity alignment suppresses weak correlations and gated fusion of modality-specific offsets enables reliable homography estimation. In the reconstruction stage, masks and attention are employed to integrate multi-view features and generate seamless stitched images. Experimental results demonstrate that the proposed method outperforms existing multispectral stitching methods in both visual quality and quantitative performance.

Zitieren

APA

Lu, X., Niu, C., & Hao, M. (2026). Multispectral infrared–visible image stitching based on transformer architecture and feature complementarity. Applied Optics, 65(21), 7028. https://doi.org/10.1364/ao.600053

Vancouver

Lu X, Niu C, Hao M. Multispectral infrared–visible image stitching based on transformer architecture and feature complementarity. Applied Optics. 2026;65(21):7028. doi:10.1364/ao.600053

BibTeX
@article{Lu2026,
  author = {Lu, Xue and Niu, Chunhui and Hao, Ming},
  title = {Multispectral infrared–visible image stitching based on transformer architecture and feature complementarity},
  journal = {Applied Optics},
  year = {2026},
  volume = {65},
  number = {21},
  pages = {7028},
  doi = {10.1364/ao.600053}
}
Kostenlos starten →