基于点云模型的低质汉字骨架提取算法

Algorithm of skeletonization of low- quality character based upon point cloud model

低质汉字的骨架提取是骨架提取中的一个困难问题.在多种降质因素的影响下,传统骨架提取方法很难提取出"好"的骨架,本文提出利用点云模型提取低质汉字的骨架.点云模型不仅能够充分利用现有汉字的底层信息,也能够将低质汉字骨架提取转化成一个两步的优化问题.采用增量广义均值聚类方法提取出低质汉字的初始骨架;然后基于高层马可夫随机模型连接初始骨架.实验结果表明,本方法在多种降质因素影响的情况下也能够获得"好"的汉字骨架.

Skeletonization of low -quality Chinese character （ LCC ） is a difficult problem . Since a variety of low -quality factors make traditional model cannot work properly .A novel model for LCC that is named point cloud model （ PCM ） was proposed in this paper .PCM can make full use of the existing underlying information of LCC , and the skeletonization of LCC was solved by a two-steps optimal problem.The primary skeleton segments （PSSs） of LCC were extracted based upon incremental generalized k-means clustering algorithm .The PSSs were combined within the framework of high -level Markov Model （ HMM ） .Experi-ments demonstratec the proposed method can generate “good” skeletons even in scenarios degraded with various disturbances .