基于版式边缘增强算法的倾斜检测

Skew Document Image Detection Based on Edge Enhancement Algorithms

该文针对数字化文档图像的倾斜现象,首先定义了多种变换算子,然后在倾斜检测中,利用这些算子对预处理后的二值化倾斜图像进行变换处理。所得到的变换图像淡化了原文档图像的内部细节特征,同时清晰地反映了原图像版面边界的走势。变换图像通过一个图像块的形式直接表征了原版面的边界信息,增强了版式边缘,使得文档图像的倾斜角度更易检测。最后通过选取有代表性的倾斜版面图像,对该算法进行了实验验证分析,得到了版面倾斜检测的统计结果。

Several transform operators are defined to deal with the skew of digital document image. In the procedure, pretreated binary skew image is processed by these operators. The transformed image ignores the detailed feature contained in the original image, while the direction tendency of layout edge is well presented. The original edge information is represented by the transformed image block. The layout edge is enhanced in the image block, and this makes it easy to detect the skew degree of original image. Typical layout images are processed by this algorithm, and the statistical result of skew detection is also presented.