Abstract: In older to solve the problem of high time complexity of exemplar-based image inpainting algorithm, and according to the characteristics of mural image, an adaptive mural image inpainting algorithm based on content is presented. By analyzing the image local changes in the gradient, the blank regions are classified into flat regions and texture regions. The texture regions are repairing with new exemplar-based image inpainting algorithm, and the flat regions are filled-in with the improved Fast Marching Method （FMM）. Finally, a general inpainting algorithm is developed. The result shows that the algorithm obtains the high efficiency.

Key words: ancient murals, image inpainting, texture synthesis, Fast Marching Method（FMM）

摘要： 为了解决基于样本图像修复算法时间复杂度高的缺点，结合唐墓室壁画的特点，提出了一种基于内容自适应的唐墓室壁画修复算法。该算法通过分析图像局部梯度变化将受损区域分为平坦区域和纹理区域，对于纹理区域采用改进的基于样本图像修复算法进行修复，而对平坦区域采用改进的基于快速行进算法（FMM）进行填充，最后提出自适应修补算法。实验结果表明，该算法在保证图像修复质量的同时提高了算法的效率。

关键词: 古代壁画, 图像修复, 纹理合成, 快速行进算法（FMM）