手写字符轮廓曲率的特征提取和识别

Handwritten Character Recognition & Features Extraction from the Curvatures of Contour Points Based on Wavelet Transform

首先对手写体字符图像进行平滑 ,同时获取字符的轮廓点 .然后提出了一个字符轮廓曲率计算的有效方法 ,为了利用小波变换对曲率信号进行去噪和平滑 ,该方法不是直接从轮廓点出发 ,而是进一步抽样得出新的轮廓点 .最后得到了不受字符旋转、平移、大小、位置影响的字符特征 ,以及相应的识别算法 .实验结果表明 。

The images of handwritten characters are smoothed and its contour points are obtained. In order to de noise and smooth the signal of the curvatures by the wavelet transform, an effective computer method of the curvatures of contour points is proposed which isn't directly based on the contour points but on the sampling. The character features that are unchanged by its rotating, parallel moving, size and its position transforming and corresponding recognition algorithm are got. Experimental results show that the method is superior to other methods while only one sort of features is adopted.