A U-Net-Based Approach for Fundus Retinal Image Segmentation

In this paper, the problem of fundus retinal image segmentation is investigated. The U-Net which is based on convolutional neural networks is used to explore the fundus retinal image segmentation. The U -Net structure includes a contraction path and an expansion path, and the U-Net can effectively combine the context information and position information of an image to achieve accurate segmentation. The data augmentation techniques including gray scale conversion, standardization, histogram equalization, and gamma transformation are employed to preprocess images to enhance the image features and improve the segmentation accuracy. The experimental results on the DRIVE dataset show that the proposed U-Net-based approach achieves excellent segmentation performance with an accuracy of as high as 95.6%.