Diffusion Autoencoder for Unsupervised Artifact Restoration in Handheld Fundus Images

2026-04-17

AI Summary

Handheld fundus images are frequently degraded by unstructured artifacts such as specular reflections, uneven illumination, and motion blur, which severely compromise diagnostic analysis. This work proposes an unsupervised diffusion autoencoder that integrates contextual encoding with a denoising mechanism to effectively restore handheld images corrupted by complex artifacts, using only high-quality desktop fundus images for training. Notably, the method operates without paired data or predefined artifact models and represents the first application of diffusion autoencoders to unsupervised handheld fundus image restoration, substantially enhancing robustness to unstructured degradations. Experimental results demonstrate that diagnostic accuracy based on the restored images reaches 81.17% across diverse artifact types and unseen datasets, with both quantitative metrics and qualitative assessments confirming the superiority of the proposed approach.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Non-convex Optimization

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The advent of handheld fundus imaging devices has made ophthalmologic diagnosis and disease screening more accessible, efficient, and cost-effective. However, images captured from these setups often suffer from artifacts such as flash reflections, exposure variations, and motion-induced blur, which degrade image quality and hinder downstream analysis. While generative models have been effective in image restoration, most depend on paired supervision or predefined artifact structures, making them less adaptable to unstructured degradations commonly observed in handheld fundus images. To address this, we propose an unsupervised diffusion autoencoder that integrates a context encoder with the denoising process to learn semantically meaningful representations for artifact restoration. The model is trained only on high-quality table-top fundus images and infers to restore artifact-affected handheld acquisitions. We validate the restorations through quantitative and qualitative evaluations, and have shown that diagnostic accuracy increases to 81.17% on an unseen dataset and multiple artifact conditions

Problem

• artifact restoration
• handheld fundus images
• unsupervised learning
• image degradation
• ophthalmologic imaging

Innovation

• diffusion autoencoder
• unsupervised artifact restoration
• handheld fundus imaging
• context encoder
• generative image restoration