Modeling and Enhancing Low-Quality Retinal Fundus Images

Retinal fundus images are widely used for the clinical screening and diagnosis of eye diseases. However, captured by operators with various levels experience have a large variation in quality. Low-quality increase uncertainty observation lead to risk misdiagnosis. due special optical beam imaging structure retina, natural image enhancement methods cannot be utilized directly address this. In this article, we first analyze ophthalmoscope system simulate reliable degradation major inferior-quality factors, including uneven illumination, blurring, artifacts. Then, based on model, clinically oriented network (cofe-Net) is proposed suppress global while simultaneously preserving anatomical retinal structures pathological characteristics analysis. Experiments both synthetic real demonstrate that our algorithm effectively corrects low-quality without losing details. Moreover, also show correction method can benefit medical analysis applications, e.g., vessel segmentation optic disc/cup detection.