基于特征点映射的楷书作品笔画骨架提取研究

【作者】 王建军；

【导师】 李艺；

【作者基本信息】 南京师范大学 ， 教育技术学， 2017， 硕士

【摘要】 汉字笔画提取作为计算机书法的关键问题,该问题的解决对于传统书法文化的传承以及计算机书法教学的实现有极其重要的作用。而当前对汉字笔画提取的方法较为单一,多局限于结构模板匹配法或统计方法。结合前人研究基础,本文以楷书为例,提出一种将模板匹配与统计法有机结合的汉字笔画提取方法。通过对汉字笔画提取工作的深入研究,提出了一种基于特征点映射的楷书作品骨架笔画提取方法。该方法首先提取书法专家联机书写楷书时的书写时序点集信息,以此构建映射模板。然后采用优化后的索引表法对采集到的楷书作品单字图片进行骨架提取,同时对骨架上的特征点进行采集。使用改进后的一致性点漂移算法进行模板与目标之间特征点映射,从而确定之间的对应关系,根据特征点之间的对应关系将楷书作品单字图片中包含的标准笔画提取出来。针对索引表骨架提取方法中的不足本文提出了对应的改进算法。改进后的方法利用提取骨架中存在的“毛刺”以及交接位置的“畸变”部位的几何拓扑关系,将“毛刺”和“畸变”进行定位,然后采用相应的方法予以去除和修复。使用改进后点匹配算法对提取到的特征点进行点集映射。一致性点漂移法采用统计学意义的软指派,对非刚性点集之间的映射具有良好的鲁棒性,基于汉字骨架特征点的连通性特点,将同一分支上的点归集为一类特征点,增加映射精确度。经过点映射后骨架上的特征点与模板中的点集建立一一映射关系,从而可以按照模板中点的笔画归属关系将骨架上的点进行归并为同一笔画内,完成笔画提取。实验表明,本文所提出的方法针对处理后的楷书作品笔画提取具有较好的效果,同时算法具有极强的鲁棒性。但由于楷书作品在初始预处理可能难以达到较为理想的状态,因此仍有很大提升的空间。更多还原

【Abstract】 Chinese character strokes extracted as a key problem in computer calligraphy,The solution of the problem is of great importance to the inheritance of traditional calligraphy culture and the realization of computer calligraphy teaching.Current method of extracting Chinese characters is more single,more limited to the structure of the template matching or statistical.Basis of previous research,this thesis take regular script as an example,put a method which combine template matching and statistical method for Chinese character strokes extraction.Based on the in-depth study of the extraction of Chinese characters,proposed an method which skeleton stroke extraction based on feature point mapping.First,extracts the calligraphy experts to build the mapping template when writing the regular time set of information in the regular script.Then the optimized index table method is used to extract the skeleton of the regular script,and collected the feature points on the skeleton in the same time.The improved Coherent point drift registration algorithm is used to map the feature points between the template and the target,get the correspondence between them.According to the correspondence between the feature points,the standard strokes are extracted.In view of the shortcomings in the method of extracting the skeleton of the index table,an improved algorithm is proposed.The improved method utilizes the geometric topology of the "burr" in the extracted skeleton and the "distorted" part of the handover position to locate the "burr" and "distortion" and then use the corresponding method to remove and repair it.The improved point matching algorithm is used to map the extracted feature points.The coherent point drift registration method has a statistically significant soft assignment,which has good robustness to the mapping between non-rigid point sets.Based on the connectivity features of the Chinese character skeleton feature points,the points on the same branch are grouped into a class Feature points,increase the mapping accuracy.After the point mapping,the feature points on the skeleton are set up with the set of points in the template,so that the points on the skeleton can be merged into the same stroke according to the stroke attribution of the midpoint of the template.Experiments show that the method proposed in this paper has a good effect on the stroke of the regular script,and the algorithm has strong robustness.But since the regular script in the initial pretreatment may be difficult to achieve a more ideal state,so there is still much space for improvement.更多还原