摘要
本文提出了一种基于改进网格特征的文本无关的离线笔迹识别方法。首先,对一个固定大小的网格窗口进行位置标注;接着,用该网格窗口去记录每一个边缘像素点周围满足改进网格特征生成条件的连通像素点对的位置信息;然后,统计出整幅图像中所记录的不同连通像素点对的位置信息特征;在获取特征向量之后,采用改进的加权卡方距离度量笔迹特征向量之间的相似度。在包含300个书写人的手写笔迹图像库上进行实验,实验结果证明了该方法的有效性。
An offline text-independent writer recognition method based on modified grid feature is pro-posed. Firstly,every square in the fixed grid was marked with different symbols. Seeondly,the positions of connected edge pixel pairs which satisfied the conditions of the modified grid feature were recorded by the grid around every edge pixel. Then, the probabititic density distribution of different connected pixel pairs is obtained. The improved and weighted Chi-square distance was applied to measure the similarity of the two handwriting. The experiments on a database containing 300 writers of handwriting images demonstrate the validity of the proposed method.
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2013年第3期132-137,共6页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家科技支撑计划资助项目(2011BAK05B04)
关键词
笔迹识别
改进网格特征
加权卡方距离
writer recogaition
modified grid feature
weighted Chi-square distance
