Ancient Chinese rubbings preserve invaluable historical and cultural information, yet their inscriptions often suffer from partial damage due to aging, environmental factors, and human handling. The diversity of calligraphic styles further complicates the restoration task, as it demands both structural accuracy and stylistic consistency. This paper proposes a novel image inpainting method named U-PatchGAN, which integrates a modified U-Net generator with a PatchGAN discriminator. This architecture enables precise reconstruction of missing character regions while preserving the original stylistic features of the rubbings. Extensive experiments on multiple datasets demonstrate the superior performance of our approach, achieving a PSNR of 21.17 dB and an MS-SSIM of 0.97, outperforming other tested configurations. The results highlight the effectiveness of our method in restoring the structure and visual texture of ancient Chinese characters, offering a promising solution for the digital preservation and analysis of historical documents.