As an important part of cultural heritage, the protection and display of tomb murals face the challenges of natural erosion, man-made damage and single means of display. This paper introduces the structure of different variational PDE class restoration models, and adopts the CDD model for mural painting restoration. By introducing geometric curvature function parameters to optimize the pheromone propagation mechanism and integrating geometric, physical, behavioral and rule models based on digital twin technology, it supports the dynamic evolution and multi-scenario application of mural paintings in their whole life cycle. The CDD model achieves the highest accuracy in different types of mask restoration, especially in the restoration of small masks, with a PSNR as high as 36.5265. Meanwhile, the CDD model still maintains a high level of performance in superimposed mask restoration, with a PSNR that exceeds that of the BSCB and TV models by 78.58% and 38.04%, respectively. The mean value of measurement error of scheme E (9.40m) is 49.11% lower than that of scheme A, and the standard deviation of its error distribution is also the smallest, which determines that scheme E can be selected.
