中文摘要：

对含有多种噪声的复杂背景笔迹图像的字符提取进行了研究。提取灰度共生矩阵的二次统计量惯性矩特征来进行图像分割，采用Hough变换方法来检测图像中的表格直线。笔迹图像字符提取步骤包括格线检测、图像分割、区域生长恢复字符笔画、去格线和断笔修补。实验证明了该方法的可靠性。

英文摘要：

To investigate the character extraction from complex background handwritten images, inertia quadrature, which is the second order statistic feature of gray level correlation matrix, is used for speckle reduction and foreground-background separation of document images, Hough transform is used for form frame line detection. The character extraction consists of line detection, speckle reduction, region growing, line removal, and broken stroke mending. The experimental results show that our proposed method is effective.