新津观音寺明代壁画图像的计算机自动虚拟修复研究

Research on automated virtual restoration of the Ming Dynasty murals in a Guanyin Temple in Xinjin

文物领域利用数字化文物通过网络媒介传播文物信息已经成为一种文物交流的新趋势。如何更好地利用采集的文物数字化档案,发挥数字文物的作用是进行文物数字化的最终目的。为此,针对新津观音寺明代壁画各种病害造成的不同程度的缺损状况,利用数字图像修复技术对其进行了计算机自动虚拟修复研究。研究表明,Criminisi基于样本的图像修复算法可利用图像小波域的小波系数优化因子,保证算法对不同纹理程度的图像采用适当的优化因子值以实现修复策略的调整,其采样的"素材"同实际壁画修复一样来自于画面已知区域,在观音寺壁画数字图像上有较好的虚拟修复效果。

It is becoming more and more popular in the cultural heritage field to disseminate cultural relic information using digital media. One of the important aims of digitizing cultural heritage data is to better use the archived digital data to enable the data to fulfill their roles. Using image technology based on analysis,we have proposed a new automated virtual restoration protocol for the Ming Dynasty murals of a Guanyin Temple in Xinjin. We found that the Criminisi algorithm,which is based on the exemplar image restoring model and which uses wavelet energy factors to guide the filling order,can restore information about missing parts very well. It adjusts the restoration rules using appropriate factors for images with different textures. The filling texture patches are from the remaining parts of the same mural,just as with real restoration methods.