<p>Cracks are one of the most common pavement surface diseases. Long-term non-maintenance can lead to crack lengthening and expansion, increasing maintenance costs. Therefore, timely crack detection is essential for ensuring pavement safety and serviceability. However, accurate crack detection remains challenging due to complex background textures, varying illumination conditions, and the presence of thin and low-contrast cracks. To address this issue, an edge-enhanced U-shaped network, named EEU-Net, is proposed for pavement crack segmentation. First, an edge enhancement module (EEM) is introduced to enhance the fine-grained details for better performance in capturing edge information. Then, an implicit attention upsampler (IAU) is presented to recover high-frequency structural details through coordinate-aware implicit representation and attention-driven refinement, thereby achieving more accurate crack reconstruction. Finally, a cross-spatial feature fusion gate (CSFFG) is introduced to integrate encoder-decoder features through a cross-spatial gating mechanism, which dynamically adjusts to accentuate vital features and selectively focuses on relevant regions. Sufficient experimental results on three publicly available datasets-Crack500, DeepCrack, and CFD-demonstrate that the proposed method achieves state-of-the-art performance, with F1-scores of 0.7930, 0.8530, and 0.7341, and mIoU values of 0.8169, 0.8673, and 0.7837, respectively. The proposed method serves as an efficient and cost-effective solution for automated pavement crack detection, facilitating the assessment and maintenance of road surface quality.</p>