Virtual Restoration of Cracks in Murals Based on Maximum Entropy Thresholding and Planar Structure Information

Abstract. Ancient murals, as invaluable cultural heritage, carry profound historical and cultural significance. However, they are often subjected to environmental factors such as temperature fluctuations, humidity, and seismic activity, which frequently result in the formation of surface cracks. This study proposes an image processing-based approach for the digital conservation of murals, introducing a method that integrates maximum entropy thresholding with planar structure information to virtually restore crack degradation in high-resolution mural images. The proposed technique comprises two main stages: crack extraction and restoration. Crack regions are identified using a combination of bottom-hat transformation, maximum entropy thresholding, connected component labeling, and mathematical morphology. Following extraction, a planar structure information algorithm is applied to restore the affected areas. By leveraging the inherent planar characteristics and structural regularities of mural images, these features are incorporated as constraints within the PatchMatch algorithm to improve restoration results. The approach has been successfully applied to real mural cases. Compared to existing techniques, it not only produces superior visual outcomes but also shows clear advantages in quantitative metrics such as SSIM, PSNR, and AGD. This virtual restoration method effectively and authentically eliminates cracks in high-resolution mural images, providing valuable guidance for practical restoration efforts while preserving the integrity and artistic essence of the murals. Thus, it contributes to the advancement of high-definition digital conservation and the long-term protection of cultural heritage.