Retinex-based enhancement of fundus images

When dealing with fundus medical images, due to low visibility, confirming the diagnosis of ocular diseases makes it difficult to conclude, and at the same time, the low-quality input image brings certain difficulties to the subsequent processing work, such as image segmentation. To solve this problem, this paper proposes a low light image enhancement (LIME) method based on Retinex theory. Firstly, the initial illumination map of each pixel is estimated by calculating the maximum value of R, G, and B channels. Finally, the final illumination map is obtained by fine-tuning the initial illumination map by introducing a priori structure on the initial illumination map. With the finely constructed illumination map, image enhancement can be performed effectively. Compared with existing enhancement methods, the method proposed in this paper achieves superior results in terms of objective assessment and enhancement quality.