Research Papers research paper arxiv computer-vision image-recognition

Decoupling Wavelet Sub-bands for Single Source Domain Generalization in Fundus Image Segmentation

arXivMarch 31, 20262 min read0 views

arXiv:2603.28463v1 Announce Type: new Abstract: Domain generalization in fundus imaging is challenging due to variations in acquisition conditions across devices and clinical settings. The inability to adapt to these variations causes performance degradation on unseen domains for deep learning models. Besides, obtaining annotated data across domains is often expensive and privacy constraints restricts their availability. Although single-source domain generalization (SDG) offers a realistic solution to this problem, the existing approaches frequently fail to capture anatomical topology or decou — Shramana Dey, Varun Ajith, Abhirup Banerjee, Sushmita Mitra

View PDF HTML (experimental)

Abstract:Domain generalization in fundus imaging is challenging due to variations in acquisition conditions across devices and clinical settings. The inability to adapt to these variations causes performance degradation on unseen domains for deep learning models. Besides, obtaining annotated data across domains is often expensive and privacy constraints restricts their availability. Although single-source domain generalization (SDG) offers a realistic solution to this problem, the existing approaches frequently fail to capture anatomical topology or decouple appearance from anatomical features. This research introduces WaveSDG, a new wavelet-guided segmentation network for SDG. It decouples anatomical structure from domain-specific appearance through a wavelet sub-band decomposition. A novel Wavelet-based Invariant Structure Extraction and Refinement (WISER) module is proposed to process encoder features by leveraging distinct semantic roles of each wavelet sub-band. The module refines low-frequency components to anchor global anatomy, while selectively enhancing directional edges and suppressing noise within the high-frequency sub-bands. Extensive ablation studies validate the effectiveness of the WISER module and its decoupling strategy. Our evaluations on optic cup and optic disc segmentation across one source and five unseen target datasets show that WaveSDG consistently outperforms seven state-of-the-art methods. Notably, it achieves the best balanced Dice score and lowest 95th percentile Hausdorff distance with reduced variance, indicating improved accuracy, robustness, and cross-domain stability.

Subjects:

Computer Vision and Pattern Recognition (cs.CV)

Cite as: arXiv:2603.28463 [cs.CV]

(or arXiv:2603.28463v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2603.28463

arXiv-issued DOI via DataCite (pending registration)