文档处理中背景字符的去除

Removing of Preprinted Characters in Document Image Processing

识别域图像的提取是文档自动处理系统中一个重要的预处理过程。在实际应用中,用户填写的信息常常与版面中的框线和背景字符存在交叠现象,严重影响了系统的性能。本文提出了基于点边距离分析的背景字符去除算法。首先通过灰度图像匹配的方法精定位背景字符子图像;然后利用形态学方法结合笔画的宽度信息对背景字符子图像进行二值化;最后分析像素点到边界距离的变化确定需要填充的像素位置,并通过形态学方法计算像素的填充值。实验采用了真实票据图像中的日期域,实验结果表明本文的方法获得了基本令人满意的效果,背景字符像素被成功去除。

Extraction of recognition item is an important preprocess procedure in a Document image analysis system. In reality, user fill-in data usually cross or touch the preprinted lines and characters, creating tremendous problems for the recognition engines. In this paper, we proposed a practical preprinted character removing method. Image matching algorithm is applied to locate the position of the preprinted character, and then the character image is binarized by mathematical morphology method combing with stroke width information. Last, the preprinted character is removed based on the varying of stroke contours. Experiment results on real-life check images demonstrate the efficient of the proposed method.