数字壁画病害提取与修复算法研究

Research on disease extraction and inpainting algorithm of digital grotto murals

数字图像修复为古代石窟壁画的科学保护与修复提供了重要依据,目前数字壁画病害信息的完整提取与合理化修复是文物虚拟修复的重要课题。以麦积山数字壁画为研究对象,分析风化裂隙伴随点状剥落的病害特征与形成机理,病害信息的完整提取以文物修复的最小干预为原则,采用以加权平均为相似性度量的局域最优层次聚类进行提取,并形成病害信息掩膜。以此掩膜图像为对象,其合理化修复利用增加了有效像素估计的曲率驱动扩散模型实现图像修复。该方法使得麦积山石窟壁画病害的虚拟修复更精确,达到了文物修复最小干预原则的效果。

Digital image inpainting provides an important basis for the scientific protection and restoration of ancient cave murals.At present,the complete extraction and reasonable restoration for disease regions of digital murals is an important topic of virtual restoration of cultural relics.This paper took the mural image of Maijishan Grottoes as the research object.The method analyzed characteristics and formation mechanism of crack diseases accompanied by point spalling firstly.Then the complete extraction of disease information was based on the principle of minimum intervention of cultural relic restoration.It used the local optimal hierarchical clustering with weighted average as similarity measurement to extract the disease information mask.Taking the mask image as the object,the reasonable restoration used curvature driven diffusion model with effective pixel estimation to achieve image restoration.This method makes the virtual restoration of Maijishan Grottoes more accurate and achieves the effect of minimum intervention principle.