Image Segmentation In The Document Image Processing Applications

With the development of information technology, document images are widely used in the projects of OA (Office Automation), DA (Digital Library), EC (Electronic Commerce), electronic government, etc. In these applications, paper documents are usually scanned into digital document images for storage and data processing. Informationization in paper documents is one of the important ways to improve the efficiency of data management and query efficiency.The document image processing includes: image preprocessing, image segmentation, image recognition. Here we focus on the image segmentation of the handwritten documents, separating the text block of images from natural images to achieve the line segmentation and character segmentation.Based on the document image texture and structural characteristics, first we do the research on the document image preprocessing with the use of HOUGH transform to detect, position the trip information and tilt angle of document for the image correction. And then we make use of the connected domain algorithm to mark the target of the image, and with the use of the structural characteristics and demographic characteristics of document image we get the image segmentation and Plain text regional.When it comes to the character segmentation we use the methods including a vertical projection, connected domain, and the structural features analysis of characters. We can get the conclusion of the effectiveness and robustness of the algorithm in this paper from experimental data.Our works are as follows:(1) Analyzing the texture features of the natural images and text files image and showing the differences between the documents images and nature images with the quantitative indicators.(2) Doing the analysis and comparison for the document image preprocessing while showing the flow chart of the image preprocessing algorithm according to the characteristics of document images.(3) Proposing a new document image segmentation method, dealing image edge with connected domain and getting the target's statistical characteristics to realize the image segmentation and presenting the specific algorithms.(4) Integrating varieties method for characters segmentation with better practicality.(5) Pointing out the shortcomings of the algorithm, making the improvement recommendations, and prospecting the further studies.