To achieve precise segmentation of optic disc and optic cup in fundus images, we designed an end-to-end joint segmentation method based on the SCUNet++ framework. This method combines the strengths of Transformer and CNN, while also leveraging the classic U-Net structure for segmentation. The SCUNet++ model was trained using publicly available fundus databases REFUGE and ORIGA. The method was tested on 370 fundus images. The Dice coefficient for disc segmentation was above 0.90 on both datasets, while for cup segmentation, it was slightly lower but still above 0.86. The test results demonstrate that this method can accurately segment the optic disc and optic cup in both normal fundus images and images with pathological retinas, indicating good applicability and generalization capability.