基于最小二乘法拟合的汉字图像笔画点研究

摘要

结合C程序，将预处理后的单个汉字图像与原图像进行逐个像素对比以判断读写，描写出原汉字字形。对经过预处理的单个汉字图像进行分析，提出了运用最小二乘法对二值化笔画点进行分组拟合的方法，从分布散乱的像素点中拟合出直线或曲线，画出汉字笔画，并计算相关系数、相关指数、残差及其平方和等参数，评估相关性、回归特性等拟合效果。最后，采用计算坐标平均的方法平整左右上下线，将其矫正成左右边竖直、上下边水平的口字形。%With C programming, the paper proposes a single Character image after the pre-image and the original character pixel-by-pixel to determine reading and writing, and to describe the original Chinese Characters. Almost all single character pre- processing image are analysed. Least square method is proposed to binary grouping strokes point fitting method. From the distribu- tion of scattered pixels in a straight line or curve fitting, it draws the character strokes and calculate the correlation coefficient, cor- relation index, residuals and their squares, and other parameters, assesses the correlation and regression fitting effect characteristics.