The task of face image restoration can be achieved through the image-to-image translation. An improved face image restoration model is proposed based on the typical pix2pix framework in this paper. This model introduces perceptual loss and style loss on the pix2pix framework to enhance the generator's ability to handle image details and global consistency. Secondly, we integrate residual blocks in the network implementation process of the model to alleviate gradient explosion and increase the stability of the model. The experimental results show that the improved pix2pix model achieves better visual performance, with significant improvements in objective evaluation metrics such as PSNR and SSIM. These results demonstrate the effectiveness of the proposed model and provide a viable solution for the face image restoration task.