基于画作线条结构分解的高清古画修复  

Repairing High-Definition Ancient Paintings Based on Decomposition of Curves

作　　者：马伟 龙晴晴 秦悦 徐士彪[2] 张晓鹏[2] Ma Wei;Long Qmgqing;Qin Yue;Xu Shibiao;Zhang Xiaopeng(Faculty of Information Technology;Beijing University of Technolog;Beijing 100124）2.National Laboratory of Pattern Recognition,Institute of Automation,Chinese Academy of Sciences,Beijin)

机构地区：[1]北京工业大学信息学部,北京100124 [2]中国科学院自动化研究所模式识别国家重点实验室,北京100190

出　　处：《计算机辅助设计与图形学学报》2018年第9期1652-1661,共10页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61771026;61379096);北京市自然科学基金(4152006);模式识别国家重点实验室开放课题基金

摘　　要：高清中国古画包含纹路纵横的画布和线条交错的画作内容,画面结构复杂.为了实现自然的古画修复效果,提出一种交互式引导的分解式修复方法.首先分离画作内容和画布.之后,结合张量投票算法和用户交互线索对画作线条分解并逐条修复.同时采用基于样例的修复方法填补画布.最后整合修复后的画作内容和画布.以多幅中国古画高清图像为实验数据验证该方法,并与Laplacian修复等方法进行修复效果的主观和量化比较.实验结果表明,该方法修复的区域与周边衔接更加自然.High-definition ancient paintings have complex structures composed of interlaced drawing curves and distinct canvas textures. Aiming at restoring high-definition ancient Chinese paintings naturally, this paper presented an interactive repairing method based on decomposition of drawing curves. First, a painting was parsed into contents and canvases. Next, our method decomposed the contents into individual curves and repaired the curves one by one, based on Tensor Voting and limited user assistance. In the meanwhile, we adopted an exemplar-based method to restore canvases. At last, we merged the repaired contents and canvases. Taking multiple high-definition images of ancient Chinese paintings as experimental data, we val- idated the proposed method and compared its results with those obtained by state-of-the-art methods, e.g. Laplacian inpainting, subjectively and quantitatively. Experimental results show that the proposed method performs better in natural completion of high-definition ancient Chinese paintings.

关 键 词：高清古画修复 线条分解 张量投票 中国古画修复 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]