Abstarct: Abstract Accurate segmentation of blood vessels in retinal images is of great importance due to its crucial role in the early detection of ocular diseases such as diabetic retinopathy and glaucoma. This study presents an effective approach for automatic blood vessel segmentation baxsed on an improved U-Net deep learning architecture. The proposed model employs CLAHE preprocessing and data augmentation techniques, including rotation and zooming, to enhance segmentation accuracy. Furthermore, convolutional attention blocks and compression mechanisms are integrated to extract both spatial and channel-wise features effectively. Experimental results demonstrate that the model achieves a high accuracy of 0.9523, outperforming existing methods. The experiments were conducted on the Retina Blood Vessel Segmentation dataset. The proposed model can serve as a powerful tool for automated ocular disease detection in clinical applications.