灰度文档图像字符切分方法

目的解决在较低质量灰度图像中进行字符切分问题．方法提出了一种字符切分方法，先对图像预处理来消除灰度不均匀性。再通过各个灰度级下的所有连通域的伪边界总长与面积的平方根之比找出一个阚值。使用泛滥填充算法，找出每个小于等于闽值像素的八连通域的边界，完成字符块预切分，最后找出重叠块和粘连块完成切分．结果通过对超过200幅的低质量灰度图像进行实验，获得了较满意的结果．结论由不同灰度级下连通域的伪边界总长与面积平方根的比值与连通域分散程度的相关性，可确定灰度图像中字符切分所用的阚值，经过合适的预处理可以在很大程度上消除低质量灰度图像对切分的影响．

A Study on Character Segmentation for Gray Document Image

To solve character segmentation for low-quality gray image,the paper presents a method for character segmentation.First,grayscale asymmetry is removed by image pretreatment.Secondly,the threshold value is found out by the ratio of artificial-borderline length all of interconnected domain and square root of area.Thirdly,all the eight-interconnected domain borderline is found out,which is less than threshold value pixels or amount by flood-fill algorithm,and then character block is finished beforehand segmentation.At last segmentation is realized by finding out all overlap block and conglutination block.The relative satisfactory result is achieved by experiments of more than 200 low-quality gray images.The ratio of that different grade of grayscale interconnected domain artificial-borderline length and area square root are relevant which separates degree of interconnected domain and can be used to determine threshold value that influences character segmentation of gray image.