HCDFormer: A Hybrid Convolution and Dimension-based Self-attention Method for Image Motion Deblurring
International Journal of Computational Intelligence Systems · 06.04.2026 · 19 (1)
DOI: 10.1007/s44196-026-01293-0 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computational Intelligence Systems, online erschienen am 06.04.2026. Verfasst von Weizhi Yang und 4 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Guangdong Key Discipline Scientific Research Capability Improvement Project, Scientific Projects of Guangzhou Xinhua University, Plan of Fifth Doctoral Guidance of Guangzhou Xinhua University.
Abstract (Originalsprache)
Convolution-based neural networks have been extensively employed in motion deblurring. However, two issues prevent them from realizing their full potential: (1) limited receptive fields owing to the kernel size; (2) failure to adequately characterize the complex and varying degrees of blur in the inference stage, owing to static weights. Although self-attention can alleviate these issues, its computational complexity is prohibitively high, as it scales with the square of the length of the image token sequence. Therefore, we propose a hybrid convolution and dimension-based self-attention method for motion deblurring, which efficiently combines the advantages of convolution and self-attention. First, a dimension-based self-attention module focusing on the macro-scale blur patterns from row, column, and channel dimensions is introduced to compute long-range attention at low computational cost. Second, a DL-Conv block consisting of depth-wise separable convolution with a large kernel size and layer normalization is employed to extract micro-scale local blur features with a larger receptive field. The complementarity between micro-scale local blur features and macro-scale blur patterns enables comprehensive characterization of the blur information. Finally, a complementary loss function that considers different aspects, including pixel-level loss, frequency-domain loss, and structural similarity loss is adopted to train the model. Extensive experimental results demonstrate that the proposed model outperforms other state-of-the-art (SOTA) methods. Compared with other self-attention-based methods, the proposed model achieves better performance with fewer parameters and shorter inference time.
Förderung
- Guangdong Key Discipline Scientific Research Capability Improvement Project
- Scientific Projects of Guangzhou Xinhua University
- Plan of Fifth Doctoral Guidance of Guangzhou Xinhua University
- Projects of Young Innovative Talents in Universities of Guangdong Province
- 2023 Guangdong Undergraduate University Teaching Quality and Teaching Reform Project
Zitieren
APA
Yang, W., Xie, Y., Ni, J., Gao, Y., & Tang, J. (2026). HCDFormer: A Hybrid Convolution and Dimension-based Self-attention Method for Image Motion Deblurring. International Journal of Computational Intelligence Systems, 19(1). https://doi.org/10.1007/s44196-026-01293-0
Vancouver
Yang W, Xie Y, Ni J, Gao Y, Tang J. HCDFormer: A Hybrid Convolution and Dimension-based Self-attention Method for Image Motion Deblurring. International Journal of Computational Intelligence Systems. 2026;19(1). doi:10.1007/s44196-026-01293-0
BibTeX
@article{Yang2026,
author = {Yang, Weizhi and Xie, Yi and Ni, Jiangqun and Gao, Ying and Tang, Jun},
title = {HCDFormer: A Hybrid Convolution and Dimension-based Self-attention Method for Image Motion Deblurring},
journal = {International Journal of Computational Intelligence Systems},
year = {2026},
volume = {19},
number = {1},
doi = {10.1007/s44196-026-01293-0}
}
EUVIMED