Perspective Correction Method for Chinese Document Images

Perspective distortion often appears in the image documents which were token by digital camera. This phenomenon will lead to recognition errors or failures. Therefore, a new correction algorithm is proposed in this paper for perspective distortion images of Chinese documents. The algorithm makes use of Chinese document image´s horizontal characteristics of text line and Chinese characterpsilas features of vertical strokes, to find distortion information and then rectify the perspective image. This method does not require information of imagepsilas edge and paragraphpsilas format. It has a good effect against the incomplete perspective images and irregular paragraph´s format. Experiment show that it takes on fast, accurate and high-robust features when using this method to correct perspective distortion in the document images.