为了对字符图像进行稳定有效的细化和线段特征提取，提出了一种同时完成字符图像线段提取和细化的算法。在定义字符图像的线段区域、交点区域和子段区域，同时指出待细化字符图像是线段区域的集合的基础上，该算法首先基于图像行(列)连通像素段来提取字符图像中所有的交点区域和子段区域，然后对属于同一线段区域的所有子段区域和交点区域进行合并，最后用Bernstein—Bezier多项式方法拟合每个线段区域的骨架。该算法的鲁棒性分析及实验结果表明，该算法可取得稳定有效的细化及线段提取效果。

This paper presents an algorithm for extracting line segments from character images and thinning character images simultaneously. At first, the line segment region, the subsegment region and the junction region of an image are defined, and an image to be thinned is considered as a set of line segment regions. Secondly, themethod of extracting all junction regions and sub2segment regions is described based on connected pixel segments in rows and columns of an image, and themethod ofmerging all subsegment regions and junction regions that belong to the same line segment region is presented. The Bernstein Bezier polynomialmethod to fit every line segment region’s skeleton is also discussed. Finally stability analysis and experimental results of the proposed algorithm are reported.