Skew Detection For Form Document Using Vertex-chain-code

Nowadays document image processing using computers has become more and more useful in different kinds of applications. Form is a familiar document format. With the characteristic of simpleness and normativeness, form is a highly refined and concentrative method for information, which is being widely applied in all aspects of national economy and daily life because its function which is accurate, perspicuous, formal and easy to fill in and so on. To input, store, analyze and manage the forms by computers automatically have become an important part of intelligent document processing.When we get an image from a scanner, the image usually appears more or less skew. The skew may not only make the further analysis and the incision of the words more difficult, but also have a bad effect on the accuracy in the character recognition. Accordingly, skew detection and skew correction of images for text documents and form documents is the first step for form recognition and image analysis and the precondition of layout analysis, word pick-up and OCR of the form as well. The accuracy rate of skew correction will directly take effect on whether or not we manage to pick up the information.To the questions that the present algorithms are of great amount of computational complexity or working not so well against the heavy noise, we bring forward a brand-new detecting algorithm about images' angles of skew forms, taking the advantage of chain code, which is easy to get the coordinate of corner. The algorithm is simple and efficient, and moreover, it is robust against heavy noise.At the end of the paper, experimental data under different noise conditions are given, which on the one hand proves the correctness and validity of the algorithm and on the other hand also indicates the weak side.