Application Research On De-Skew Technique For Scanned Document Images

In recent years, with the rapid development of information technology, online scoring, electronic library and other office automation are more widely applied, which are usually require to convert huge amounts of paper documents into electronic documents. In the process of conversion, the influence of walking paper mechanism of scanner or the existence of some human factors, the images of scanned often have certain angles tilt. However, in document image processing, most of the OCR (Optical image Recognition) system is very sensitive to the skew of document images, the quality of skew correction direct impact on the quality of the image processing of the follow-up work, therefore the research on skew correction for scanned document images has a important significance.In order to correct the skew angle of document images, the first step is to detect the skew angle. This paper discussed the current skew angle detection algorithm for document images, make a brief introduction to the advantages and disadvantages of various algorithms and application occasions; Meanwhile, introduced several image preprocessing methods, such as image binarization, image denoising and image enhancement, etc.; Presented a new de-skew algorithm for scanned document images which fusion block copy, the neighboring interpolation, double bilinear interpolation and cubic spline interpolation algorithm. Using unsharp masking stressed the sharpness of images and analyzed the experiment results. The experimental results show that the algorithm has high efficiency and assure the quality of the image after rectification.Innovation points in the paper are mainly embodied in the following aspects:(1) Fast skew angle estimation algorithm:Edge point detection and fast Hough transform that limits the angle;(2) Block classification algorithm:According to the smooth degree of image region, divided into smooth block and non-smooth block;(3) Pixel level classification:Divided into smooth pixels, clarity emphasize pixels and non-smooth pixels (the pixels which need high quality interpolation)according to the first-order neighborhood maximum differential image;(4) Clarity emphasize:Unsharp masking clarity emphasizes method that combined with the pixel level classification;...