中文摘要：

快速并行算法（FPA）是经典算法中比较好的一种细化算法，但其存在固有缺陷，可能产生信息丢失。为了克服这个缺点，首先验证了FPA映像算法，在此基础上，提出了FPA重合算法，最后为了减少计算时间和冗余信息，提出了基于模板保留的FPA算法。实验结果表明，基于模板保留的FPA算法运行速度快，对手写体汉字能很好地得到中心骨架，没有断点，较好地保持了原图像的拓扑结构，且细化为只有一个像素宽的图像。

英文摘要：

Fast parallel algorithm（FPA） is one of the best classical thinning algorithms for binary-value images. However it has some inherent shortages and may generate information missing. To overcome this drawback,firstly, FPA mapping algorithm is verified. Based on this, FPA superposition algorithm is given. Finally, a fast parallel thinning algorithm based on reserved templates is proposed to reduce the computational time and redundant information. The experimental results show that the novel algorithm is fast. It can get the central skeleton without discontinuous points,keep the shape of the original images perfectly,and get the skeleton of only one-pixel width.