DEF-Net: A Dual-Encoder Fusion Network for Fundus Retinal Vessel Segmentation

The deterioration of numerous eye diseases is highly related to the fundus retinal structures, so automatic vessel segmentation serves as an essential stage for efficient detection eye-related lesions in clinical practice. Segmentation methods based on encode-decode structures exhibit great potential tasks, but have limited feature representation ability. In addition, they don’t effectively consider information at multiple scales when performing fusion, resulting low fusion efficiency. this paper, a newly model, named DEF-Net, designed segment vessels automatically, which consists dual-encoder unit and decoder unit. Fused with recurrent network convolution network, proposed, builds convolutional branch extract detailed features accumulate contextual features, it could obtain richer compared single structure. Furthermore, exploit useful scales, multi-scale block used facilitating efficiency designed. Extensive experiments been undertaken demonstrate performance our proposed DEF-Net.