Stroke Extraction Of Chinese Calligraphy Characters Based On Topology Graph

The strokes extraction of Chinese calligraphy characters has always been an important research content in the field of computer calligraphy.In previous studies,many mature algorithms have been proposed to extract simple calligraphy characters strokes.However,for some relatively complex calligraphy characters in the number and structure of strokes,there hasn't been relatively perfect algorithm.In order to establish a computer calligraphy evaluation system to evaluate the difference between self-written characters(self-written calligraphy characters)and post characters,this paper proposes a stroke extraction algorithm based on the topological structure of known stroke calligraphy characters(online writing or acquisition from standard fonts).The algorithm mainly includes the following processes:Firstly,the pre-processing of known strokes calligraphy images and self-written calligraphy characters images is carried out:denoising,binarization and morphological optimization.Then,the boundary tracking algorithm is used to extract the boundary of the calligraphy characters in the image.Then,we use the PBOD(point-to-boundary orientation distance)algorithm to calculate the distance from each pixel in the boundary of the calligraphy character to the boundary of the calligraphy character in the 360 degree direction,and the PBOD curve is drawn.According to the number of peaks in the curve,different areas of the stroke are determined,namely the endpoint area,the common point area and the intersection area.Then,the end and intersection regions of the known stroke calligraphy characters are obtained by doing the above operations.These areas are numbered according to stroke order,and the smallest endpoint area is selected as the starting point to search for the known stroke calligraphy characters in Breadth First Search.The tree topology and directed graph of known stroke calligraphy characters are established by the adjacency and relative position relations of each feature area of known stroke calligraphy characters in the traversal process.Then,the same method is used to build tree topological graph and directed graph for self-handwritten characters.By comparison,the feature regions(i.e.endpoint regions and intersection regions)of known stroke calligraphy characters are corresponded one by one to the feature regions of the original calligraphy characters.Finally,the tree topological graph is used to search the stroke path representing the known stroke calligraphy characters,and the path is corresponded to the calligraphy characters to extract the strokes.By randomly selecting calligraphy characters for stroke extraction experiments,it is proved that the stroke extraction algorithm based on topological graph can correctly extract calligraphy strokes.