Ancient Wall Painting Restoration Based on Improved Adaptive Sample Block Criminisi Algorithm and CDD Algorithm

To better protect and restore ancient mural art, this study proposes an improved adaptive sample block Criminisi algorithm to address the shortcomings of traditional manual restoration methods. To enhance the efficiency and quality of image restoration, this study constructs a curvature driven diffusion model based on curvature adaptive parameters and smoothing functions. The experiment showed that compared with the Criminisi and the enhanced Criminisi algorithms, the average PSNR of the research algorithm increased by 12.47% and 7.17%, and the structural similarity index increased by 5.54% and 2.82%, respectively. The subjective average score of this algorithm for image sample restoration was 9.1 points, and the average time for image restoration was 23.25 seconds. Compared with deep residual network and generative adversarial network, the proposed algorithm increases the peak signal-to-noise ratio by 36.59% and 20.42%, respectively, and reduces the image restoration time by 31.92% and 24.80%, respectively. The results indicate that the proposed image restoration method can improve the quality and efficiency of image restoration. This study can provide certain reference value for artificial restoration and has important research significance for the continuation of world history and culture.