EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

DMSNet: Dual‐Branch Modulation Selective Network for Image Denoising

Xiaotao Shao, Xuanming Hu, Yan Shen, Zhenghang Chen

IET Image Processing · 2026-01 · 20 (1)

DOI: 10.1049/ipr2.70440 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Image Processing, online erschienen am 21.08.2026. Verfasst von Xiaotao Shao und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Despite the remarkable progress of transformer‐based architectures in image restoration, existing approaches typically entangle spatial and channel modelling within a unified architecture, which is sub‐optimal for real‐world denoising where noise is spatially variant and signal‐dependent. In this paper, we propose DMSNet, a dual‐branch modulation selective network that explicitly decouples spatial and channel representations to better exploit their complementary roles in image denoising. Specifically, the channel branch employs a hybrid modulated channel attention module to perform noise‐aware feature calibration via scale‐shift modulation, enabling robust global dependency modelling. Meanwhile, the spatial branch introduces overlapped window spatial attention to break the boundary constraints of conventional window attention and facilitate seamless cross‐window information propagation. In addition, the outputs of the two branches are finally integrated by a global selective fusion block, which adaptively balances spatial and channel cues according to global contextual information. Extensive experiments on both synthetic and real‐world benchmarks demonstrate that DMSNet achieves competitive performance while maintaining favourable computational efficiency.

Zitieren

APA

Shao, X., Hu, X., Shen, Y., & Chen, Z. (2026). DMSNet: Dual‐Branch Modulation Selective Network for Image Denoising. IET Image Processing, 20(1). https://doi.org/10.1049/ipr2.70440

Vancouver

Shao X, Hu X, Shen Y, Chen Z. DMSNet: Dual‐Branch Modulation Selective Network for Image Denoising. IET Image Processing. 2026;20(1). doi:10.1049/ipr2.70440

BibTeX
@article{Shao2026,
  author = {Shao, Xiaotao and Hu, Xuanming and Shen, Yan and Chen, Zhenghang},
  title = {DMSNet: Dual‐Branch Modulation Selective Network for Image Denoising},
  journal = {IET Image Processing},
  year = {2026},
  volume = {20},
  number = {1},
  doi = {10.1049/ipr2.70440}
}
Kostenlos starten →