Research on Mural Restoration Method Based on Generative Multi-column Transformer
Hui Ren, Fanhua Zhao, Zhen Li, Ke Sun, Shilin Gao · 2022 IEEE 5th Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC) · 2022
Ancient mural are splendid artistic treasures of ancient civilizations, but their blurring and destruction are becoming more and more serious due to the effects of time and environment. In this paper, a Generative Multi-column Transformer method is proposed to recover the missing areas of mural. The method synthesizes different image components in a parallel manner by transformer and enhances local details by employing implicit diversified Markov random fields regularization. Generative Multi-column Transformer is used in combination with reconstruction and Markov random field loss methods to propagate local and global information from the background to the target defaced region to reconstruct the loss in that part. The experimental results show that the method based on Generative Multi-column Transformer adopted in this paper is able to restore mural images with complex textures and missing areas very well, while the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) of the method are improved, this paper provides theoretical guidance for the digital restoration of murals.