课题以高清采集的墓室壁画残存信息为对象，研究被病害污染区域画面信息重建的数字修复技术。遵循格式塔视觉心理学原则，充分考虑信息重建过程中的病态性，通过提取主体信息、颜色信息、结构与纹理信息等先验知识，利用先验知识引导信息扩散和填充到病害污染区域；自动识标记龟裂、裂隙与断裂等病害区域，产生掩膜信息，建立数据模型；改进CDD扩散方式、优化优先值计算方法，改进样本取样方式，增加结构约束合成建立相应的修复模型，分别满足小尺度信息缺失、大尺度信息缺失、样本冗余与扩展、结构中断的分块拼接等四个方面的修复需求。并搭建在线交互式数字修复方案论证平台，吸取远程专家经验，论证数字修复效果，降低直接手工修复壁画可能产生的风险，解决壁画修复任务多与各博物馆修复人员紧缺的矛盾，缩短壁画从出保管库到进展厅的周期。

This project research the technique on the tomb mural inpainting for museum. According to the Gestalt principle of visual psychology, and take the ill-posed question in to condition. By analyzing the left information of the those murals, it has established digital archives and divided the restoration into three parts as building the prior acknowledge, digital model and inpainting model. The prior acknowledge include the subject and background information, the color layout, cartoon information and textural information. They can help to choose the exemplars from the original mural exactly; the digital model include to mask the parts which those diseases as crack, crevice and fracture ruined, and let the computer know where need inpainting; the inpainting model include improved CDD model for small scale filling, improved exemplar-based model for big scale filling, the reduce redundant for high definition data, the structural constraints for splicing divided blocks. And build an tomb mural inpainting platform for the evolution and get the suggests from cultural relics experts to adjust the strategy of inpainting technique. The project will help lower the risks by restore the mural directly, and solve contradictions between the heavy task and few restorers, and shorten the restoration period.