As an important heritage of ancient Chinese art, murals not only carry rich historical information but also reflect the cultural characteristics of various periods. The current mural restoration methods have a high degree of damage and unsatisfactory restoration results. In response to this, this study proposes a new method for mural protection and digital restoration. Firstly, the multi-scale Retinex algorithm is studied based on adaptive multi-scale guided filtering and colour restoration factors, achieving mural image enhancement. After image enhancement, a network model combining transformer model and U-shape network is applied to extract and segment the features of mural images. Finally, digital restoration of murals is achieved by combining channel and spatial attention mechanisms. The experiment showed that the expert score of the repaired image using the research method reached 4.52 points, and the peak signal-to-noise ratio and structural similarity were 24.58 dB and 0.94. In practical application, the qualified rate of mural image restoration using the research method reached 96.5%, and 115 murals could be restored per hour. Therefore, the proposed digital restoration method can effectively improve the efficiency and accuracy of mural restoration.
Pei Liu (Mon,) studied this question.