Slant Correction of Document Image Based on Line Structural Analysis
Xiao Li · 2007
Document digitalizing provides a method to utilize the archival resources of government comprehensively. Slant correction of digitalized document is necessary for the quality of digital image, and it’s a preprocessing of any other procedure. By a Gaussian filter of multi-scale, the global structure of the original document image was enhanced. Based on gradients of enhanced image, a structural analysis tensor is also defined for the detection of line structure at the each pixel. For the line-likeness value at every pixel, which is greater than a given number, the corresponding angle is calculated. From the histogram of these angles, document skew can be determined. Our experiments show that the proposed method is adaptive to various documents with rich text.