数字壁画病害提取与修复算法研究.

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. [ABSTRACT FROM AUTHOR]