Structure-Consistent Restoration Network for Cataract Fundus Image Enhancement

Fundus photography is a routine examination in clinics to diagnose and monitor ocular diseases. However, for cataract patients, the fundus image always suffers quality degradation caused by clouding lens. The prevents reliable diagnosis ophthalmologists or computer-aided systems. To improve certainty clinical diagnosis, restoration algorithms have been proposed enhance of images. Unfortunately, challenges remain deployment these algorithms, such as collecting sufficient training data preserving retinal structures. In this paper, circumvent strict requirement, structure-consistent network (SCR-Net) images developed from synthesized that shares an identical structure. A simulation model firstly designed collect sets (SCS) formed sharing Then high-frequency components (HFCs) are extracted SCS constrain structure consistency preservation SCR-Net enforced. experiments demonstrate effectiveness comparison with state-of-the-art methods follow-up applications. code available at https://github.com/liamheng/Annotation-free-Fundus-Image-Enhancement .