基于轮廓追踪的字符识别特征提取

字符识别是模式识别的一个重要分支，其关键是特征向量的选择与提取。小波分解和分形在图像处理方面有着广泛的应用，在结合二者特点的基础上提出了一种新的基于轮廓追踪的字符识别特征选取方法。即对于一个输入的字符图像经预处理提取其轮廓，并由轮廓追踪获得边缘点坐标序列，实现了从二维图像数据到一维数据的转化，对得到的一维曲线进行小波分解，计算少数几个分解得到的曲线的分形维数，以它们构成特征向量。并对有关字符做了实验，其效果是令人满意的。

Feature extraction in character recognition based on contour pursuit

Character recognition is an important branch of pattern recognition,its key factors are selecting and extracting proper feature vector.Wavelet decomposition and fractal are applied extensively in image processing,and a new feature vector combined their characteristics is proposed in this paper based on contour pursuit.After an input image is preprocessed,contour is extracted.Then edge pixels coordinate sequence is obtained based on it.This method transforms 2-D image data into 1-D data which is decomposed by wavelet to get curves.Afterwards the feature vector is formed by calculating fractal dimension of several segments of curves.Some characters are tested using the method,and the result is satisfied.