Pre-Processing Of Document-Image With Complex Layout

With more and more application of computer in information field, especially the development of Internet Technology, Electronic Document has been the most important carrier nowadays. People have paid great attentions to the digital processing of documents. In any document processing system, pre-processing always plays a pivotal role which affects all the later modules in the system. Because most of other modules in the system only deal with binary image , Gray-level image to Binary- image conversion, namely Image Binarization, is a key process in the course of pre-processing.Image Binarization is the separation of object and background, and in terms of document image, it means the division between text and background. But as the diversification of typeset and development of printed technology, the document image has been more and more rich and colorful, which challenges the system of document image processing.In thesis, we will first introduce an integrate Document-Processing System and its Preprocessing module. Then, we will emphatically give attention to the problem of Binarization, and make a survey of various existing image-thresholding methodologies, including global methods and local dynamic methods, by evaluating their excellences and shortcomings. Next, specially in the case of document image with complex gray-level variation, we will offer an improved dynamic thresholding algorithm, and prove its feasibility by experimental simulation. Finally, we will discuss the problems possibly encountered in application systems, and put forward the basic principle of question resolving in those cases.