Abstract

Low-quality fundus images caused by blur, abnormal exposure, occlusion, and illumination artifacts can reduce the reliability of clinical interpretation and intelligent diagnosis. To address this problem, this paper proposes FreqTrans-IQA, a fundus image quality assessment framework based on frequency-domain degradation transfer and high-frequency structure perception. First, high-frequency response point maps and one-way Chamfer distance are used to construct structurally matched high- and low-quality image pairs, reducing the interference caused by anatomical layout differences during degradation transfer. Then, a degradation frequency encoder and a frequency-domain transfer generator are designed to transfer degradation-related amplitude-spectrum information from real low-quality images to high-quality images, while a Patch discriminator is introduced to constrain the local realism of the generated degraded samples. In addition, an RGB-HFC dual-branch Swin Transformer is employed to jointly model spatial semantic information and local high-frequency degradation details for fundus image quality prediction. Experimental results on a public quality grading task and a private degradation category classification task show that the proposed method can effectively improve the discrimination ability of the model for fundus image quality grades and degradation types, demonstrating the effectiveness of frequency-domain degradation modeling and high-frequency structure perception.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/OMIA_025.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to Open Review

Open Review Page: https://openreview.net/forum?id=2cMM0qD1uX

BibTex

@InProceedings{XueLiy_Fundus_MICCAISAT2026,
        author = { Xue, Liyi AND Liu, Ping AND Chen, Yinshan AND Zhao, Xinyu AND Wu, Zhenquan AND Zhang, Guoming AND Lei, Baiying AND Wang, Tianfu},
        title = { { Fundus Image Quality Assessment Based on Frequency-Domain Degradation Transfer and Generative Adversarial Networks } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17270},
        month = {pending},
        page = {pending}
}


back to top