基于全局和局部特征的中文笔迹鉴别方法研究

提出一种基于全局和局部特征的中文笔迹鉴别方法.利用一种改进的Gabor变换提取笔迹图像的全局特征,对得到的特征集进行聚类后,笔迹集由于书写风格不同被分成有效和无效两个子集,在有效的集合中通过矩法进行局部特征提取,最后采用欧氏距离进行分类.对笔迹库的实验表明,提出的方法得到了满意的识别率.

Integration of Global and Local Feature for Writer Identification of Chinese Handwriting

A method combining the global and local features was presented for writer identification of Chinese handwriting. A new Gabor filter was used to extract global features from handwritten documents. The features were divided into two classes by k-means clustering according to the difference of writing style. Then, moment method was used to extract local features in valid class and Euclidean distance classifier was adopted to compare the writings. Experiments showed this method can get promising results.